Modeling of Next Generation Digital Learning Environments: Complex Systems Theory

Modeling of Next Generation Digital Learning Environments

Series Editor Fabrice Papy

Modeling of Next Generation Digital Learning Environments Complex Systems Theory

Marc Trestini

First published 2018 in Great Britain and the United States by ISTE Ltd and John Wiley & Sons, Inc.

Apart from any fair dealing for the purposes of research or private study, or criticism or review, as permitted under the Copyright, Designs and Patents Act 1988, this publication may only be reproduced, stored or transmitted, in any form or by any means, with the prior permission in writing of the publishers, or in the case of reprographic reproduction in accordance with the terms and licenses issued by the CLA. Enquiries concerning reproduction outside these terms should be sent to the publishers at the undermentioned address: ISTE Ltd 27-37 St George’s Road London SW19 4EU UK

John Wiley & Sons, Inc. 111 River Street Hoboken, NJ 07030 USA

www.iste.co.uk

www.wiley.com

© ISTE Ltd 2018 The rights of Marc Trestini to be identified as the author of this work have been asserted by him in accordance with the Copyright, Designs and Patents Act 1988. Library of Congress Control Number: 2018954298 British Library Cataloguing-in-Publication Data A CIP record for this book is available from the British Library ISBN 978-1-78630-316-5

Contents

Foreword . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

ix

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

xi

Chapter 1. A Virtual Learning Environment seen as a System of Instrumented Activities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 1.1. From school radios to MOOCs: a retrospective glance at the evolution of instrumented activities in education . . . . . . . . . . . . . . . . . . . . . . . . . 1.1.1. Educational technologies, ICT, ICTs for teaching, common ICTs? . . 1.1.2. A broad variety of technologies in education . . . . . . . . . . . . . . . 1.1.3. Learning to put knowledge, expertise and interpersonal skills into practice. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.1.4. Learning modes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.1.5. Learning modes in education . . . . . . . . . . . . . . . . . . . . . . . 1.1.6. ICT, learning and pedagogical theories: an interrelated revolution . . . 1.1.7. From objectivist epistemology to school radio and television . . . . . 1.1.8. From behaviorism to computer-assisted learning . . . . . . . . . . . . 1.1.9. From construstivism to microworlds . . . . . . . . . . . . . . . . . . . 1.1.10. From social constructivism to CLE, CSCL and cMOOC . . . . . . . 1.1.11. Learning in an open network: from connectivism to MOOCs . . . . . 1.1.12. Synthesis of these evolutions . . . . . . . . . . . . . . . . . . . . . . . 1.2. VLEs: a system of instrumented activity . . . . . . . . . . . . . . . . . . . 1.2.1. Virtual learning environments (VLEs) . . . . . . . . . . . . . . . . . . 1.2.2. The principles of the systemic paradigm . . . . . . . . . . . . . . . . . 1.2.3. The steps involved in the systemic approach . . . . . . . . . . . . . . . 1.2.4. VLEs seen as open systems . . . . . . . . . . . . . . . . . . . . . . . . 1.2.5. VLEs seen as systems of instrumented activity . . . . . . . . . . . . . 1.3. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . .

1 6 8 9 11 11 12 13 15 19 23 27 38 38 38 40 44 51 55 61

vi

Modeling of Next Generation Digital Learning Environments

Chapter 2. Modeling Instrumented Activity at the Heart of the Virtual Environment. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2. Modeling instrumented activity, yes, but why? . . . . . . . . . . . . . . . . 2.2.1. What type of model are we talking about? . . . . . . . . . . . . . . . . 2.2.2. Models for multiple uses . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2.3. Models with specific uses . . . . . . . . . . . . . . . . . . . . . . . . . 2.3. Contour, components and hierarchical levels of a system of instrumented activity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.3.1. Perimeters, objects and components. . . . . . . . . . . . . . . . . . . . 2.3.2. Three levels (macro, meso, micro) associated with business processes 2.3.3. An example of distance learning device analysis . . . . . . . . . . . . 2.3.4. Associated levels, objects and models . . . . . . . . . . . . . . . . . . 2.4. Summary of models and modeling languages . . . . . . . . . . . . . . . . . 2.4.1. Models of pedagogical engineering (EML) . . . . . . . . . . . . . . . . 2.4.2. Training engineering models . . . . . . . . . . . . . . . . . . . . . . . . 2.4.3. Adaptive models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.4.4. Systemic models of activity . . . . . . . . . . . . . . . . . . . . . . . . 2.4.5. Systemic models of complexity . . . . . . . . . . . . . . . . . . . . . .

63

. . . . .

. . . . .

63 65 65 66 68

. . . . . . . . . . .

. . . . . . . . . . .

76 76 77 79 80 81 81 96 100 103 106

Chapter 3. Models of Instrumented Activity Challenged by Technopedagogical Innovations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109 3.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2. The Vygotskien model and its expansion . . . . . . . . . . . . . . . 3.2.1. Digital ink: towards a new virtual learning environment design 3.2.2. Use of tablets (iPads) in school contexts . . . . . . . . . . . . . 3.3. Expansion of Engeström’s model . . . . . . . . . . . . . . . . . . . 3.3.1. Looking for a unifying model in long-distance learning. . . . . 3.3.2. First expansion of Engeström’s model for designing a CLE . . 3.4. New context of usage and new expansion . . . . . . . . . . . . . . . 3.4.1. Digital workspaces in schools: online text books . . . . . . . . 3.4.2. Second expansion of the model: facilitating the analysis of online textbooks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.5. Expansion of pedagogical and training engineering models . . . . . 3.5.1. Resistance to pedagogical engineering models . . . . . . . . . . 3.5.2. Evolution towards training engineering models . . . . . . . . . 3.6. MOOC models to build . . . . . . . . . . . . . . . . . . . . . . . . . 3.7. Conclusion on the resistance conditions of virtual learning environment models . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . .

. . . . . . . . .

. . . . . . . . .

. . . . . . . . .

. . . . . . . . .

. . . . . . . . .

109 111 111 114 116 116 120 123 123

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

124 125 126 130 131

. . . . . .

132

Contents

Chapter 4. The Digital Learning Environment in the Paradigm of Systemic Complexity Modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

vii

135

4.1. Complex systems theory and theoretical framework . . . . . . . . . . . . . . . 135 4.1.1. Definition of a complex system . . . . . . . . . . . . . . . . . . . . . . . . 137 4.1.2. Modeling of complex systems . . . . . . . . . . . . . . . . . . . . . . . . . 148 4.2. Argument in favor of the application of the complex systems theory to the modeling of latest-generation DLEs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159 4.2.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159 4.2.2. DLE: a complex system . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162 4.2.3. Complexity theory: at the heart of the connectivist learning design . . . . 168 4.2.4. Conceptual analogy between the SM of complexity and SM of activity . . 170 4.2.5. Modeling these environments with suitable languages . . . . . . . . . . . 172 4.2.6. Enrolling in a forward-looking approach . . . . . . . . . . . . . . . . . . . 178 Chapter 5. Modeling a DLE Perceived as a Complex System . . . . . . . . . 5.1. Finalized and recursive processes in an active environment . . . . . . . . . . 5.1.1. Identifiable finalized processes (functions) . . . . . . . . . . . . . . . . . 5.1.2. Recursive processes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.1.3. Representation of the processes and functional levels of the system . . . 5.1.4. The eight constituent functions of a DLE applied to systemic modeling. 5.2. Synchronic processes (the system performs) . . . . . . . . . . . . . . . . . . 5.2.1. Identifying the active presence of interrelationships . . . . . . . . . . . . 5.2.2. Establishing an input and output value inventory for each process . . . . 5.2.3. Representing data flow and functional dependencies between processes 5.3. Diachronic processes (the system becomes) . . . . . . . . . . . . . . . . . . . 5.3.1. Typical sequences (or scenarios) and even tracking . . . . . . . . . . . . 5.3.2. Traces left by events . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3.3. The states and state diagrams . . . . . . . . . . . . . . . . . . . . . . . . 5.3.4. Process descriptions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.4. A system capable of processing information and deciding its own behavior . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.5. A model based on analysis data . . . . . . . . . . . . . . . . . . . . . . . . . 5.5.1. A model constructed from analysis data . . . . . . . . . . . . . . . . . . 5.5.2. Provenance and data collection . . . . . . . . . . . . . . . . . . . . . . . 5.5.3. Data analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

181

. . . . . . . . . . . . . .

183 183 189 190 193 197 198 200 201 206 208 208 209 210

. . . . .

211 211 211 212 224

Chapter 6. Modeling and Simulation of an MOOC: Practical Point. . . . . .

227

6.1. Modeling an MOOC . . . . . . . . . 6.1.1. Package and use case diagrams 6.1.2. “Class” diagrams . . . . . . . . 6.1.3. State transition diagrams . . . . 6.1.4. Sequence diagrams . . . . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

228 231 236 239 240

viii

Modeling of Next Generation Digital Learning Environments

6.2. MOOC simulation . . . . . . . . . . . . . . . . . . . . . . . . 6.2.1. The program structure . . . . . . . . . . . . . . . . . . . 6.2.2. The script . . . . . . . . . . . . . . . . . . . . . . . . . . 6.2.3. Related files . . . . . . . . . . . . . . . . . . . . . . . . . 6.2.4. Parameters that come into play and initialization values 6.2.5. Progress in the chapters . . . . . . . . . . . . . . . . . . 6.3. Simulation result . . . . . . . . . . . . . . . . . . . . . . . . . 6.3.1. Level of total understanding . . . . . . . . . . . . . . . . 6.3.2. Level of understanding of the module. . . . . . . . . . . 6.3.3. Level of understanding of the chapter . . . . . . . . . . . 6.4. Putting the obtained results into perspective . . . . . . . . .

. . . . . . . . . . .

. . . . . . . . . . .

. . . . . . . . . . .

. . . . . . . . . . .

. . . . . . . . . . .

. . . . . . . . . . .

. . . . . . . . . . .

. . . . . . . . . . .

. . . . . . . . . . .

. . . . . . . . . . .

243 243 245 246 246 246 246 247 248 250 251

Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

253

Appendices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

257

Appendix 1. Functional model notation . . . . . . . . . . . . . . . . . . . . . . .

259

Appendix 2. Dynamic model notation . . . . . . . . . . . . . . . . . . . . . . . .

261

Appendix 3. MOOC.py and Quiz.py . . . . . . . . . . . . . . . . . . . . . . . . . .

263

Appendix 4. Étudiants.py . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

265

Appendix 5. Chapitre.py . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

267

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

269

Index. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

289

Foreword

Understanding latest generation virtual environments such as connectivist MOOCs in all their complexity is an ambitious but essential project in today’s world. This project is so essential because these virtual learning environments (VLE), as Marc Trestini rightly says, are at the heart of the knowledge economy present today in our daily lives and exist, in this respect, in some of our universities. However, at the moment, they are rarely fully understood. How can knowledge be built, exchanged, shared and spread throughout the heart of these VLEs? To find an answer to this question, Marc Trestini has put forward the idea of adopting a paradigm for the modeling of complex systems as proposed by Le Moigne [LEM 99]. The purpose of this accessible book is to cause the reader to understand this paradigm by presenting a method, a language and an illustration of its implementation, but especially to show the reader how to “think together” rather than separately. Before embarking on this predictive approach, the author proposes a history which allows us to understand that modeling has always been indispensable for describing and understanding VLEs throughout their history, from Skinner’s teaching machine to the most modern learning environments. To approach the types of modeling adapted to VLEs, based on education technology, the author reminds us of the activity model proposed by Engeström, well adapted to the old VLEs which were limited to a small number of learners and closed by intentions previously set by the designers.

x

Modeling of Next Generation Digital Learning Environments

Marc Trestini shows that while this model is no longer suitable for the next generation of environments, which are open to a very large number of learners, he proposes a modeling approach which aims to help us to grasp the complexity of next generation VLEs. Furthermore, readers tired of empty words – which are unfortunately used more and more frequently when discussing virtual environments in teaching and learning – will find here the opportunity to learn about the heart of these VLEs in this book. They will also learn about the central phenomenon of emergent learning, marking it as the realization of instrumented activities and grasping its complexity. Finally, in reading this book, the reader will also learn a lot of essential knowledge, such as: what is modeling? What are the modeling languages of ITS (Intelligent Computing)? What is systematic modeling of complexity? How can it be implemented? Finally, what are the prospects for future research? Hope remains that this knowledge will be shared and built on by many specialists in our universities and colleges and that it will create many opportunities for research and development. Bernadette CHARLIER

Introduction

“Algorithmic reasoning is a checklist preventing forgetfulness, Heuristic reasoning is an intelligent thought. Can we not then hold the algorithms for heuristics among others, rather than as producers of certain results? We remain the only ones responsible for the results of our reasoning, even if they are computerized.1”

With the Internet, communication opportunities have continued to grow and diversify, in private or family contexts. The ease of use of Web 2.0 and its invaluable potential in terms of the exchange and sharing of information between social actors have allowed the emergence of a new society known as knowledge society. The social web (another name given to Web 2.0) represents a turning point in the organization of the bonds holding communities together and promotes the emergence of new learning communities. Training and distance learning professionals speak of social learning when they refer to spontaneously constituted communities in view of sharing knowledge, and also in the case of more formal distance learning devices that incorporate learning groups formed with less forethought2. In both cases, learning occurs most often during so-called collaborative activities. “Learning” results from peer-to-peer exchanges (socio-cognitive conflicts) and imitation through the observation of the behavior of other people who are part of the learning group. Communities that form spontaneously, where learning generally occurs outside the traditional academic or university context, are referred to as informal learning. It most often corresponds to a need for personal and relatively immediate learning that does not require any particular framework or educational structure. The underlying 1 Jean-Louis Le Moigne discussing this book. 2 When the designer of the formation decides on the formation of social groups.

xii

Modeling of Next Generation Digital Learning Environments

concept involves sharing information and knowledge on learning networks (e.g. on the free encyclopedia Wikipedia). As soon as users have a question about a word, a geographical location or the composition of a recipe, they access the information they were looking for on the Web, usually using a search engine. In this case, learning by means of these networks is beneficial due to many possibilities offered by Web technologies. These possibilities are in constant evolution and offer Internet users a variety of possible ways to access knowledge. In the second case, when the learning group is formed randomly, the learning responds to a more formalized procedure. The individual’s desire to learn through these networks remains the same, but it is also accompanied by the desire to belong to a formal learning group, that is, to be part of an institutional training structure. This type of education differs from the face-to-face method (i.e. physical presence) only in its use of networks. Signing up is generally required (and often, paying tuition fees) and the contents and the pace of the training are defined beforehand. To increase its efficiency and restore the social bond that distance tends to weaken, the designers of this type of training most often choose to introduce a collaborative dimension to learning. This pedagogical modality, which is not specific to long-distance learning, is now particularly effective in this context. By these different means, both technological and educational, long-distance learning offers new opportunities to train those who live in remote areas, who work or who have a busy schedule. The formal framework of traditional training is preserved, while offering more adaptability and flexibility to registered users thanks to the numerous possibilities offered by the networks. There are also many other training opportunities using networks and an infinity of nuances between formal and informal conceptions of e-learning. Depending on how the device is designed, the subtle combination between these two dualistic visions can be balanced anyhow and these learning methods are then characterized as semi-formal. We will have the opportunity to return to this topic in detail as virtual learning environments based on this hybrid design are the subject of this book. We refer to them as the latest generation virtual learning environments (VLEs). Khan Academy, MIT OpenCourseWare (OCW) and even Massive Open Online Courses (MOOCs) are very good examples. Khan Academy offers exercises, videos and an adaptive learning platform that allows users to learn at their own pace. MIT OpenCourseWare is a project providing, online and free of charge, the entire educational resources used by MIT for its university courses. MOOCs and their derivatives allow users to learn thanks to online courses throughout the world. The first MOOC was launched in 2008 in Canada.

Introduction

xiii

The idea of its founders was to put an existing face-to-face university course (CCK083) online and to open it for free to everyone regardless of age, qualification or nationality. No upper bound to the number of users was then defined. At the beginning of this experiment, the founders had registered 25 face-to-face students, but after opening this course online, they had 23,000 more. These devices, more or less informal, are revolutionizing the way we learn and teach today. Their effects on the economy and on the national policies are beginning to be felt. Our economies are increasingly based on the flow and sharing of knowledge and information. Hence, they are often referred to as knowledge economies. Indicators of this economy include knowledge development and management, such as research and development (R&D) expenditures, the employment rate of skilled workers, and the intensity of the use of new information and communication technologies (NICT). The OECD defines knowledge-based economies as “those that are directly based on the production, distribution and use of knowledge and information”. In fact, like many other next generation VLEs, MOOCs are a major contributor to the development of these economies and will most likely play a key role in the years to come. Among the different latest-generation virtual learning environments, MOOCs occupy an important place in this book. As we have just described, they are, of course, perfect examples of new-generation learning environments, but more generally, they are also essential because they are about to revolutionize learning processes and all levels, from primary school to the university. Analyses that focus on these particular environments can be equally applicable to other distance learning devices as long as they are, like MOOCs, devices that are open to all and that do not impose any conditions of age, qualification or nationality, and that do not set a maximum number of users. These characteristics best express the idea we have of the latest generation VLEs. If we are to believe that these environments are really about to revolutionize the world of education and training (as the media frequently announce), it will be harder for us to predict what they will be tomorrow and what kind of novelties they will bring. It is also for this reason that we believe that modeling can be useful not only to evaluate their impact on learning, but also to predict some of their socio-economic effects in the medium and long term. To take into account their undeniable complexity, in this book, we propose to study them with tools provided by the “complex system modeling” paradigm [LEM 99]. The principles underlying this modeling, also called systemic modeling (SM), seem to us to be more adapted to the analysis of these environments than those of traditional analytical modeling (AM). The general principle of the latter consists only of simplifying observed phenomena by operating on simple disjunctions and reductions. If SM also relies on the need to distinguish and 3 CCK08: The Connectivism and Connective Knowledge course began in 2008 (hence the 08), taught by Stephen Downes and George Siemens.

xiv

Modeling of Next Generation Digital Learning Environments

analyze, it “seeks more to establish the communication between what is distinguished, the object of the environment, the thing observed and its observer. It does not try to sacrifice the whole thing for one part, or one part for the whole thing, but to conceive the difficult problem of the organization” [MOR 90, p. 43]. It is therefore by applying the principles of complex system modeling to the analysis of the latest generation VLEs that we believe we can properly analyze these environments and observe possible emerging phenomena, or even identify some important trends that may be useful to different economic, political and social actors around the world. Models also facilitate the development of knowledge about certain phenomena observed, such as the learning processes involved. More generally, the simulation methods related to this systemic modeling of complexity can predict possible evolutions that can sometimes approach the system’s real behavior. Using these methods, it becomes possible to predict possible solutions by trying a very large number of initial conditions, and also to observe what happens if we change certain interactions or certain parameters. Before addressing these modeling questions, which are at the heart of this book, we propose to review the historical evolution of digital learning environments in education in order to better understand the technical and pedagogical evolutions that these environments have undergone and of which the most recent ones have benefited. In fact, since the early work of Skinner and his very famous “teaching machine”, technologies have not ceased to bring to the world of education their share of novelties. Today, we realize that technologies have not fundamentally changed the way we learn4, but that they have been able to serve the various pedagogies chosen by the teachers themselves. They therefore act as a catalyst by allowing us, in the best case, to accelerate or improve certain relatively well-known learning processes. Throughout the last century, educational technologies and cognitive psychology have not evolved independently of each other but rather jointly, the former helping to “energize” the latter. First-generation technologies, for example, merely spread knowledge and responded perfectly to the transmissive conception of the teaching of the time. From Gutenberg’s printing press to educational radio, the aim was to convey knowledge to as many people as possible. The techniques then implemented were based on a transmissive mode of learning in which the teacher gives their knowledge to the learner, who was ostensibly lacking the knowledge in question. Then came Skinner’s behaviorist theories that aim to bestow a lasting behavior on the learner. To do this, the students’ responses, in line with those expected, were systematically followed by positive reinforcements (good marks, congratulations, encouragement, right to continue the exercises, etc.). Skinner himself was the first to create a technology that served this theory, which he named the teaching machine. 4 In the sense of cognitive processes and not teaching methods.

Introduction

xv

The principle of this rudimentary machine was later reinvented when the first microcomputers appeared for use in classrooms. Specific training programs, also known as exercisers, have allowed many students to progress in their school education. They have been the subject of several experiments and have opened a whole field of research in “educational technology”: computer-assisted teaching (CAT). These technologies are an improvement on the traditional systems as they allow the student to revise or deepen certain topics without the presence of the teacher. They have also made it possible to reinforce automatic learning for specific learning programs. The answer to the question asked is immediate and repeated as much as necessary, so as to help the student memorize the correct answer. It now seems established that their use plays an important role in tutoring as well as for deepening one’s knowledge. However, behaviorism has failed to create consensus. For example, Jean Piaget was at the origin of a profound change in the way we conceive learning while Seymour Papert thought about the place the computer could have within this new paradigm. Piaget proposes an epistemological approach to learning which he calls constructivism. For him, knowledge is a mental construct of reality that is forged by confronting it. Learning must not be reduced to a transmission, but the learner must instead be an actor. Seymour Papert thought that we should no longer offer children a programmed learning, but that they should instead become the builders of their own knowledge. According to this epistemological approach to learning, the teacher must therefore make tools available to the learner, who must use the tools and become familiar with their use, as if they were his or her own. The first of these many tools is the language that the child assimilates; he/she learns writing through paper and pen, and then uses all sorts of mediation objects like the blackboard for example. Digital technologies for educational purposes can be easily designed in this framework and they are simply more complex tools which a child can learn to master. This theory has given birth to virtual learning environments using computers (VLEs). The goal is to create “microworlds”, of which one of the first was an educational programming environment using Logo, a programming language: the learner tests commands that make it possible to move a small robot called a turtle, and thus appropriates the language at the origin of the movement. Finally, Soviet theorists such as psychologist Lev Vygotsky added a social dimension to constructivism by founding the socioconstructivist movement. This says that one learns better in contact with others, because people can discuss their difficulties, help each other, watch what others do, etc. To use a famous formula of Vygotsky [VYG 97, p. 355], “what a child can do today with assistance, she will be able to do by herself tomorrow”. In fact, the instrumental and collective activity we alluded to above is at the heart of this theory. In education, this group activity traditionally takes place in the classroom, when

xvi

Modeling of Next Generation Digital Learning Environments

students engage in collaborative activities, such as carrying out a joint project or solving a situational problem together. However, it can also take place in particular digital environments called CEHL (computing environments for human learning). These environments take advantage of digital tools to promote interaction between different communities (e.g. pupils, parents and teachers) and within them. Gradually, the learner is characterized by their nodal status in a virtual group in which they interact, cooperate, exchange information synchronously or asynchronously and make decisions in an electronically constructed universe. These learning environments, by means of interposed technologies, are part of the general framework of long-distance teaching, long-distance learning or even e-learning5. When the theories of learning are seen through computer networks and connectivism, other important questions arise. For George Siemens [SIE 05], who invented this connectivist conception of learning, the natural temptation for theorists is to continue to revise and evolve the theories of learning as soon as the initial conditions are changed. For him, this is possible only up to a certain point: when the initial conditions are modified too much, the adjustment of the models to these variations becomes impossible. A completely new approach is then necessary and the questions to be explored must be dealt with in another framework: chaos theory and networks for example. This framework of analysis alone constitutes a revolution in the field of research on computing environments for human learning (CEHL). MOOCs and all their consorts are a major subject of study for these theories. For researchers in educational technologies, it is therefore a question of trying to understand, through these different theories of learning, if the use of these virtual learning environments in education and training is really relevant, as well as studying their cognitive effects. As mentioned above, this was partly made possible by the construction of models, that is, simplified representations of reality that aim to be more intelligible. Thus, following this logic, immediately after this historical reminder of the evolution of VLEs, the book explores the question of their modeling. Legendre [LEG 93] distinguished two types of models. On the one hand, object models are objects that seek to reflect as accurately as possible a reality observed using experimental data. On the other hand, theoretical models seek to reduce the observed phenomenon to a more general phenomenon. The representative models of the VLEs that we are trying to construct are mostly object models. Engeström [ENG 00, ENG 05] observed that for each new form of mediation implemented in learning environments, new models have been created and introduced to account for them. Thus, the first generation of mediation was individual mediation as described by Vygotsky. The associated model makes the tool appear as a medium between the 5 To signify their openness and flexibility in the modalities of the training.

Introduction

xvii

subject and the object during the activity. This model was adopted by Kuutti [KUU 96] to apply it to the field of learning mediated by digital tools. The second generation of mediation appeared with Engeström’s model which illustrates the collective mediation introduced by Léontiev and which accounts for the importance of the collective nature of the activity. The Subject, the Tool, the Object, the Rules, the Community and the Division of Labor are represented as well as the relations that bind them. Engeström himself builds on this model by introducing a circular model that accounts for an observation close to the Hegelian dialectic: the system of activity evolves and draws its strength from the succession of resolutions of internal contradictions appearing within the system of the activity. Thus, the act of learning is part of the transformation of the activity system. According to this theory, learning is not considered here as a mere accumulation of information, but as the reorganization of a system of activity [BRU 04, p. 3]. In this book, we consider that while the Engeström model is well suited to the analysis of traditional virtual learning environments, in particular characterized by a limited number of learners, it is much less so when it is applied to the analysis of latest generation VLEs. These latest VLEs are characterized by their great openness to the public. Indeed, the emergence of social networks, MOOCs, informal learning via networks and connectivist approaches to learning are introducing unprecedented complexity in the history of VLEs. The history is characterized in particular by the variety of learners’ profiles, enrolment motivations, age and socio-cultural categories. To study and analyze them, the current models cruelly show their limits because the “enroled” are too numerous and their profiles too different compared to those who attend traditional long-distance learning courses. Their attendance becomes difficult to “represent” in a traditional model such as that of Engeström. Their “open” character is no less difficult to model because we must take into account their openness to the whole world. Finally, their evolution seems chaotic and their effects seem scientifically unpredictable. Faced with this complexity, we began by proposing an expansion model of the Engeström model, which develops each of its poles in order to reveal the existence of certain “buried” or “masked” system components within each of them (see Figure 3.10). In this, we wanted to open “boxes” (the poles of the model) to reveal the supposed constituents that we needed to observe in order to further our analysis. We have thus embarked on splitting each of the poles of the model into clusters by establishing new relationships between each of them. A holographic model has emerged (a trundle in every way). The “tool” pole, for example, brought to light, thanks to this additional splitting, the pedagogical, didactic and technical artifacts that constitute it. This allowed us to explore the tensions that may occur between these finer constituents and which are likely to cause possible instrumental conflicts [MAR 11] or, perhaps even to ensure a form of instrumental orchestration [TRO 05]. In the same way, the splitting of the other poles of the model (object, subject, community,

xviii

Modeling of Next Generation Digital Learning Environments

rules) gave us new means of analysis. Then, during the study of an “online textbook” in a VLE school [TRE 09a], we were led to develop this holographic model in a three-dimensional space in order to represent the different working communities (parents, teachers, students) and to study the perceptible tensions between them (see Figure 3.11). We could have even gone further by adding the community of designers. It is worth noting that although the second generation of collective mediation, identified by Engeström, was the object of a formal modeling on his part, as far as we know, there is not yet a model representing the third generation of mediation which Engeström himself referred to, namely mediation at the inter-collective level. With the proposed 3D model, acting communities and the relationships that connect them have become easier to identify (see Figure 3.11). This “extended” model also facilitates the study of networks of interacting activity systems. After these undeniable advances, the models presented seemed to be able to meet the new challenges of distance education and training. Unfortunately, with the techno-pedagogical innovations of the 2010s, we realized that the growing complexity of the new generation of VLEs should lead us to continue this split, to divide infinitely the constituents/objects of the system. From a feeling of incompleteness regarding existing models came the need to break them down. However, this feeling added to the desire to move against the current and to follow instead the teachings of complex system modeling (SM): we should reason by linking together rather than by cutting to pieces. This wording is in fact the fundamental question that this book addresses. Under its systemic appearance, and in spite of all the interest of Engeström’s model and the expansions that we proposed later, was it necessary to continue this infinite division and to attach oneself to this disjunctive logic? Was it still useful to absolutely want to model a system by adopting an analytical approach that aims to study its parts separately in order to deduce a global behavior by simple summation? Was it not wiser to consider that its behavior could be more than simply the sum of the behaviors of its parts, as claimed by the SM? Although it may certainly be effective when subsystems are relatively easy to identify, the analytical modeling (AM) we then introduced was less suitable when the system became complex, because it was starting to become tedious to refine the model to a degree of detail that could quickly become staggering. Moreover, the analytical method is valid only when the subsystems are independent: otherwise, a subsystem whose operation would have been determined when it was isolated would behave quite differently once integrated into the whole system. It is clear that this approach has limits. After explaining our reflective path, in this book, we propose to adopt a completely different approach based on the paradigm of systemic modeling of complexity. This (really) systemic approach, which adopts a conjunctive approach,

Introduction

xix

considers the VLE as a complex system, fitted at each moment with a certain number of properties modifiable by its environment, but which can also act on the latter, and which transforms itself. Such a system is furthermore composed of subsystems, which are considered as fully fledged systems in the sense that has just been established, and whose environment is therefore on the one hand the other subsystems, and on the other hand the environment of the “over-system”. Thanks to this modeling, we can correctly account for the phenomena of feedback, mutual interaction between subsystems and evolution of behavioral laws over time. The obvious disadvantage is that it does not predict the effects of the system easily, and we have access to the equations governing the model but we do not necessarily know how to solve them. This disadvantage is, however, largely offset by the fact that such models (e.g. those of object-oriented modeling) make it possible to implement simulations; we can do without an analytical resolution in favor of a numerical resolution. However, the real revolution of this paradigm shift is that it allows for the first time to set up a potentially predictive model. And is not planning the future the ultimate goal of any scientific modeling? After this introduction, in Chapter 1, we describe our object of study as we perceive it, that is, as a system of instrumented activities. To explain this point of view, we show the existence of a significant link between the evolution of technologies and the evolution of conceptions of education over the last century. In particular, we show that these evolutions converge towards a conception of teaching and learning essentially centered on the realization of instrumented activities. Then, in Chapter 2, we report on the challenges of modeling by describing the objectives, the outlines and the tools it uses. This allows us to make an initial inventory of the visual models and modeling languages used in Intelligent Computing. In Chapter 3, we address the issue of the resilience of these models when confronted with a change in sociotechnical context and in particular when introducing techno-pedagogical innovations. The structural and functional modifications they cause lead us to put our work into the paradigm of systemic modeling of complexity in Chapter 4.

xx

Modeling of Next Generation Digital Learning Environments

After having described this paradigm and argued this choice, in Chapter 5, we propose a systemic modeling approach that evolves according to the nine increasing levels of complexity of a system [BOU 04a], in which each level also has all the characteristics of the lower levels. In Chapter 6, this leads us to illustrate the method recommended by applying it concretely to the modeling and simulation of a recent MOOC named “Project Management”. Before concluding and grouping together our bibliographic references, we dedicate Chapter 7 to the research perspectives that this study suggests.

1 A Virtual Learning Environment seen as a System of Instrumented Activities

1.1. From school radios to MOOCs: a retrospective glance at the evolution of instrumented activities in education “Education has always used a variety of devices, media, tools and processes to facilitate the transfer of knowledge to learners” [BAR 11, p. 109]. Of all these instruments, traditional blackboards or overhead projectors are the most wellknown. More recently, the computer, the video projector, the digital tablet, the interactive whiteboard (IWB) or distance learning devices (which MOOCs belong to) have been added to the list of frequently used instruments in education. Often referred to as “teaching aids” or “mediation tools”, both aim to publicize1 [PER 05] or mediate [FAB 13] relationships between the learner, the teacher and knowledge as it appears, for example, in Houssaye’s Pedagogical Triangle [HOU 88]. These instruments also mediate other relationships, such as those between the learners themselves, when they are led to work together. This is particularly the case for distance education platforms such as Moodle, Claroline, Sakaï, Acolad or Dokeos, which are designed to promote activities labeled “collaborative” or “cooperative”. In this type of environment, the platform and its related artifacts mediate the relationship between the different learners (the subjects) and the relationship between the learners and their learning group (their community), relationships that are also formalized in Engeström’s activity model [ENG 87, p. 78]. 1 “The point of view of mediation is that of communication, where mediation can be defined as a translation, a link between the enunciator and the receiver […] through the devices that accompany the user and facilitate the uses …” [FAB 13]. The term mediatization is also used to refer to mediation using physical instruments. But when applied to VLE, it refers instead to the work of Daniel Peraya [PER 05] who uses the term “to designate the process of creating such training and communication media, a process in which scriptwriting occupies an important place”. Modeling of Next Generation Digital Learning Environments: Complex Systems Theory, First Edition. Marc Trestini. © ISTE Ltd 2018. Published by ISTE Ltd and John Wiley & Sons, Inc.

2

Modeling of Next Generation Digital Learning Environments

In fact, so-called educational technologies, which mainly represent this group of means, can be included, as research objects, in the paradigm of human activity which considers activity as organized and spread by using instruments. According to this theory: “The actions are always inserted in a social matrix composed of individuals and artifacts […]. Thus, as the mind works through artifacts, its work cannot be bound unconditionally and exclusively to the brain or to the individual himself; it should be perceived as distributed in related artifacts. These said artifacts link individuals and actions in a permeable, changing, and eventful manner” [DUC 05]. It is also for this reason that we consider these virtual learning environments (from computer-assisted learning to MOOCs) as systems of instrumented activities falling under both the system theory and the activity theory. Their integration into the French education system began in the 1960s, and they were thus involved in the revival of schools, with the emergence of school radio and educational television programs. From 1971, the computer was then introduced into high schools to meet the requirements of computer introduction programs. This overwhelmingly established the computer in school spaces in the 1980s with the IPT2 program of 1985. This date also marks the use of the first computer networks in education (nanoarrays)3. The first exchanges of digital documents between pupils and teachers (pages of texts, images, programs, etc.) via servers have been at the source of profound changes in teaching practices. Which teacher does not use networks today to share documents with pupils? Later, the arrival of the Internet in 1971 and Web 1.0 in 1990 once again led to the discovery of new uses in education. The first information research on the Internet illustrates this renewal in the class activity. However, it was specifically due to Web 2.0 in 2002, also called the dynamic web or social web, that the “Internet” revolution was particularly marked in education. The first long-distance learning devices and the platforms upon which they rely are based on this technology and could not have developed without it. It is also that today this same Web 2.0 (with only a few evolutions) can offer the possibility of training via networks and in particular with MOOCs to anyone in the world. The latter also introduce themselves into education with their array of novelties. Since the first George Siemens and Stephen Downes’ MOOC Connectivism and Connective Knowledge course (CCK08) was launched in Canada in 2008, MOOCs have grown on an impressive scale all over the world and have become established in the educational arena, including schools and universities by supporting original pedagogical approaches. However, they have also rekindled 2 Informatique pour tous program (literally, “computing for all”). 3 The nanonetwork was designed by the Université des sciences et technologies de Lille, France in 1982. It connected a Léanord server to MO5 and TO7 Thomson computers.

A System of Instrumented Activities

3

some teaching methods such as “the flipped classroom” or the “enriched classroom”. Between 2002 and today, many other instruments have appeared in educational environments. Without being too exhaustive, it is worth mentioning the panoply of emerging technologies like the cloud (which hints at the beginning of Web 3.0), tablets, tablet PCs, smartphones, geolocation processes, augmented reality, etc. The principle of augmented reality now allows the superposition of a virtual 3D model that “floats” over a real object, a bit like a hologram, but through digital screens, for smartphones and tablets in particular. This concept, which is still emerging, is already appearing in schools. For examples, there are applications4 that make it possible to transform children’s drawings (more generally, “triggers” or “markers”) into a 3D virtual image (an “Aura”). With a simple click, the image is superimposed on the drawing by hovering the mouse over it. Azuma [AZU 97], quoted by Domingues, is one of the first researchers to define augmented reality: “Augmented Reality (AR) is a variant of Virtual Reality (VR). While virtual reality plunges an individual into a virtual environment where he/she cannot see the real world, RA allows the user to see the real world where virtual objects are superimposed on it. An augmented reality system must combine real objects with virtual objects, all in real time” [DOM 10, p. 10]. In education, many augmented reality applications for educational purposes are currently being developed. They pave the way for futuristic visions of new uses in education: in a classroom, a museum, a zoo and a library. They also concern all disciplines: language learning, geometry, science, reading and the arts [KAD 16]. Thus, the more or less explicit idea that the use of technologies in a school context brings progress and educational effectiveness has progressively developed in popular culture by reinforcing an already highly idealized collective representation of these technologies in education. However, thanks to the many research studies that have been conducted in this area, we now know that learners’ performances are not necessarily better when these technologies are used [END 12, p. 1]. In the wake of what Russell [RUS 01] called 4 “Augmented Polyhedra”: software where each augmented reality marker is associated with a geometrical figure, cube, parallelepiped, cylinder, sphere, cone, pyramid, tetrahedron, five different prisms: https://play.google.com/store/apps/details?id=com.miragestudio.polygons. “Aurasma”: a mobile application to “read” augmented reality made specifically for this application. It also allows you to create your own augmented reality. Indeed, we can easily, from a smartphone or tablet, embed an image or a video on an element of reality. With the Aurasma application, we can create, share and read “Auras”. An Aura consists of the following: a virtual key plus a trigger (belonging to reality). The trigger is what is recognized by Aurasma to trigger the embedding of the virtual element: https://www.aurasma.com.

4

Modeling of Next Generation Digital Learning Environments

the No Significant Difference Phenomenon, many people have been, and still are, critical of research carried out, deeming it inadequate for providing “proof” of the impact of technology on learning: we can recall the famous debate of 1990 between Clark [CLA 83, CLA 94, CLA 09] and Kozma [KOZ 91, KOZ 94] on the influence of media on learning. Clark did not see the media as influencing student success as the truck that delivers our food influences our diet. Kozma, for his part, recognized that technologies had so far had little influence, but that they should play a greater role in educational processes [TRE 10b]. Since this controversy, the debate continues and the questions it raises remain more relevant than ever. Philippe Dessus and Pascal Marquet [DES 09, pp. 7–10] have returned to this question in an interview with Clark himself for Distances et Savoirs magazine. While they both agree that the Clark and Kozma debate has shifted to new objects, they wonder if the fundamental questions that this debate raises have been definitively settled or if they are still being empirically tested [DES 09, pp. 7–10]. More recently, just as a vast digital educational plan for schools5 has just been announced in France, the results from the first study by the Organisation for Economic Co-operation and Development (OECD), published on 14th September 2015 on pupil’s digital skills, show that it is not enough to extensively equip students and teachers with digital tools in order to improve their performance6. For Scardigli [SCA 89] and Musso et al. [MUS 07], “Digital innovations initially generate expectations that are as excessive as they are ideological, which are typical of the technological determinism that prevails in digital discourses in education”. Taking into account the pros and cons, what does it really take to think of these virtual learning environments? Do they really help us to learn better? Are they learning tools or simply tools for learning? What is their real cognitive effectiveness? Do they effectively fulfill this role of mediation that we want them to? Do they really change the learner’s way of learning? In fact, we often hear that young people today are learning through the use of technology. Do they really learn differently, or do they always learn in the same way (i.e. according to the same cognitive processes) but using different tools, which are constantly evolving? In other words, is it the way of learning that is modified by the 5 Plan numérique pour l’éducation: 500 écoles et collèges seront connectés dès 2015 (Digital educational plan: 500 schools and colleges will be connected in 2015), http://www.education.gouv.fr/cid88712/plan-numerique-pour-l-education-500-ecoles-etcolleges-seront-connectes-des-2015.html. 6 Le Monde.fr with AFP (15 September 2015 at 18:16, updated on 15 September 2015 at 19:48), by Matteo Maillard, available at: http://www.lemonde.fr/campus/article/2015/09/15 /en-classe-le-numerique-ne-fait-pas-de-miracles_4758368_4401467.html#sp2S1RAjMfqJmlsz.99.

A System of Instrumented Activities

5

use of these technologies or is it only the teaching method and pedagogical practices that change? This question related to the effectiveness of virtual environments on learning still needs to be asked and addressed by research. While it is true that it has made good progress in this area, it has not yet found a sufficiently well-established response to overcome this kind of investigative work. This question is all the more crucial as ‒ particularly in France ‒ we find ourselves in a difficult economic situation but paradoxically favoring the integration of these technologies into education. State investment in virtual environments in schools also remains an effective economic way to boost growth. It is therefore a matter of cautiously advancing on these issues, knowing, for example, that the “Grande École du Numérique” initiative announced by former French president Hollande in May 2015 preceded the annual OECD forum, which took place a month later and discussed dealing with economic recovery. Also remember that for Angel Gurria, Secretary-General of the OECD, “traditional economic drivers are no longer enough to fuel global growth, therefore it must be oriented towards technology and skill development”7. At the same time, while these technologies are likely to boost economies, the vast equipment plans for schools in educational technologies are costly to states and therefore to taxpayers. It is therefore right to expect precise answers to these questions of cognitive (or even economic) efficiency from research. However, the question remains complex because it involves a multitude of influential factors that act between themselves and also on the system itself. A methodology that relies in particular on the paradigm of a systemic modeling of complexity seems well adapted to the situation. This is because one of its objectives is in fact to facilitate the analysis of the processes involved in this type of chaotic environment using specific tools [TRE 15b]. We will therefore deal with these questions with far more than technophile and technophobic discourse. We will specifically discuss the history of the integration of technologies into education: what motivated their integration each time? In other words, what justified their use? This retrospective look at the evolution of instrumented activities in education lends these technologies educational properties, which would be inherent to them [SEL 12]. Through the history of this integration, we will try to show that these technologies have not invented new ways of understanding, but that they have been able to put themselves at the service of pedagogies, stemming themselves from the cognitive psychologies of the 20th Century. In other words, these technologies did not influence the different cognitivist currents (i.e. how we explain and how we learn), but they have systematically adapted to their evolution. We will thus be able to conclude that ICTs 7 Challenges.fr,“Pour l’OCDE, ‘il fault investir dans la technologie et dans l’éducation’”, (“For the OECD, we must invest in technology and education), June 2, 2015, http://www.challenges.fr/economie/20150602.CHA6431/pour-l-ocde-il-faut-investir-dans-latechnologie-et-dans-l-education.html.

6

Modeling of Next Generation Digital Learning Environments

are not really cognitive tools, but that they have a cognitive potential, whose use is an important factor in the learning and skill development of the learner [DEP 09]. To demonstrate this, let us return first to the evolution of educational technologies from the second half of the last century, then, secondly, to that of cognitive psychologies during the same period, to finally cross these two perspectives and observe the influence of concomitant changes on learning processes and teaching practices. We will start with educational technologies. 1.1.1. Educational technologies, ICT, ICTs for teaching, common ICTs? Many acronyms have been and are still used to characterize these technologies. First, NICTs (new information and communication technologies), an acronym used less and less as the “N” marks the novelty which later seems so obvious that it becomes implicit. Then, there are ICTs (information and communication technologies). According to Larousse, they are defined as “the set of techniques and computer equipment for communicating remotely by electronic means”. It is only by adding the “for teaching” to ICT that these technologies are understood as occurring in a school context or “for educational purposes” (ICTT). In the same way, we sometimes also talk about educational technologies, even if this name is older. The term common ICTs (common information and communication techniques) is mainly used in French national education programs. This move from ICTs for teaching to common ICTs is not trivial; it wants to show the importance that the French Ministry of Education gives to the uses of, and not only to the control of IT tools. These uses are meant to be reasoned, responsible and relevant and all the actors of French National Education are sensitized to these questions during their training, whether at school, upper primary or junior high school through to high school and university, or when becoming teachers, through a widespread technological education. It is worth noting that since 2005, common ICTs have been the fourth pillar of the common core of competences setting the cultural and civic benchmarks that constitute the content of compulsory education in France. Since the beginning of the 2016 academic year, they have been integrated into the “methods and tools for learning” in the common or required knowledge module. It is therefore with the 1985 IPT program that ICTs started to flood the education world. The reasons which motivated the successive equipment plans are numerous. The IPT plan was intended to introduce the 11 million students of France to the

A System of Instrumented Activities

7

computer tool, and also to support the national industry (Thomson in particular). “It aimed to put in place, from September 1985, more than 120,000 machines in 50,000 schools and to train 110,000 teachers by the same deadline. Its cost was estimated at 1.8 billion francs, of which 1.5 billion was for equipment”8. Since then, equipment policies have followed one another; the last large-scale operation, before the previously mentioned “grande école du numérique”, dates back to 2009. It was a vast virtual equipment program for rural schools with a budget of 67 million euros. Since 1985, the objectives have not changed much overall, although the importance given to each of these objectives according to the socio-economic and cultural situations varies: – fighting the digital divide was a priority in 1985, but this is less and less true today as the number of households with digital devices is increasing drastically9; – preparing children for an information society, a knowledge-based economy; the school must prepare for life, so you have to “learn” about computers; – mastering the tools of current communication and the professional world, acquiring a digital culture and developing a critical attitude to the dangers of the Internet in particular. However, the objectives are also educational, and include: – effects on motivation (teacher and student); – interactivity; – ability to produce an immediate response (instant assessment); – tool for educational differentiation; – individualization of teaching; – downplaying mistakes, their status changes; – IT for disabilities (dyspraxia, motor issues, etc.); and – improved presentation of results.

8 Informatique pour tous program (2015, archived version, updated October 11): http://en.wikipedia.org/w/index.php?title=Plan_informatique_pour_tous&oldid=119398265. 9 Two out of three households have Internet in France. In ten years, the proportion of households with Internet access at home has risen from 12% to 76% in 2011. The digital divide is therefore narrowing, but differences in degree and social category remain. The computer is still the most used medium, but new ways of accessing the Internet are developing. Mobile Internet, for example, has made a significant breakthrough: 24% of Internet users accessed the web in 2010 via their mobile phone against only 9% in 2008, according to INSEE: http://www.insee.fr/fr/themes/document.asp?ref_id=ip1340.

8

Modeling of Next Generation Digital Learning Environments

1.1.2. A broad variety of technologies in education When we talk about educational technologies or ICTs for teaching, what are we referring to specifically? Is it computers only? No, of course not. The definition of ICT given above proves this. It is not because this definition does not apply specifically to the educational world that the educational world does not inherit the same tools. These terms/acronyms have also been created to avoid this confusion. There exists in the ICTs an entire panoply of tools and concepts that we will briefly recall. In education, the first technologies used were school radios in the 1930s, followed by school televisions in the 1950s. Computer science was introduced in the 1970s. The concept of multimedia made a dramatic and controversial breakthrough into education around 1990. In 1995, a number of French schools began using Internet connections. Many “educational” applications derive from it, mainly based on hypertext and hyper-landscape10 concepts. Teachers are starting to move away from traditional mass memory storage devices (CD-ROMs, DVDs, etc.) to online educational software. This period also marks the beginning of the craze for educational resources and tutorials available on the Internet. In 2002, Web 2.0 made a spectacular breakthrough worldwide, which had an impact on the world of education and training. The so-called Dynamic or Social Web (interactive on the page) allows communication between learners on platforms created for teaching or distance learning. Another form of “mediatized” socialization is gradually taking hold. Behaviors and educational practices are changing. Distance brings new elements into the act of learning and online teaching. E-learning modalities are multiplying and are the subject of careful observation by researchers in the education sciences. This is the rise of e-learning: collaborative learning platforms and online collaborative web applications (collaborative workspaces, online conceptual maps). The interactive whiteboard (IWB) has begun to appear in classrooms in France11. This device combines the advantages of a touch screen and video projection. According to some specialists, it favors metacognitive activities and would enable potential socio-cognitive conflicts that could promote learning [GIL 95]. The “PrimTICE” interactive whiteboard program has strongly encouraged primary school classes to equip themselves with this tool. Launched by the French Directorate for Information and Communication Technologies (SDTICE) in April 2004, this program allowed France to move from a dozen IWBs beforehand to around 2,500 IWBs at the end of 2007, and 14,000 were planned for the end of 2008. The Fourgous report, submitted to Luc Chatel on 15th February 2010, assesses the “Successful Digital Schools” parliamentary mission. It makes installing 10 A hyper-landscape is to the image what a hypertext is to the text. It is a medium in which it is possible to navigate through hotspots that behave like internal or external web links. 11 Smart, located in Canada, produced the first IWB in 1991.

A System of Instrumented Activities

9

IWBs in classrooms a flagship measure, “to increase the percentage of interactive whiteboards linked with a computer and software in schools to 100% in order to create multimedia educational sequences”12. Furthermore, mobile technologies (mobile phones, smartphones, tablets, tablet PCs, and podcasts) have recently been introduced into primary and secondary education, as well as into higher education. Geo-localization coupled with mobile telephony, connected objects and the cloud will certainly provide new educational applications with the development of Web 3.0. We have already previously mentioned the educational applications made possible by the concept of augmented reality; virtual visits to museums and even historic monuments in 3D are now at our fingertips. These technical developments would be nothing if they were not likely to add value to learning or teaching. To intelligently create links between ICT and learning, it is necessary to briefly return to a fundamental question: “How do we learn?” 1.1.3. Learning to put knowledge, expertise and interpersonal skills into practice What is learning? Is it acquiring knowledge, expertise or adequate behavioral skills? It is without doubt all of these. There is therefore no real consensus on the meaning of the word “learn”. For some, learning is acquiring knowledge, understanding and skills. Understanding would be for them related to knowledge, skills linked to expertise and aptitude for interpersonal skills. However, if we consider that learning is a little bit of each of these, we would say that to learn is to acquire skills that are intrinsically made up of understanding (or knowledge), abilities (or potentials) and skills (or expertise). This is also the scientific meaning given by Philippe Perrenoud, professor of education and an expert on the subject of expertise. For him, “expertise” is the ability to mobilize a set of cognitive resources (knowledge, abilities, information, skills, etc.) to face a variety of situations with relevance and effectiveness [PER 00]. In addition to being clear, this definition also has the advantage of bringing notions together which, when together, truly convey meaning. Let us go back for a moment to the three of them: knowledge, understanding and abilities, which are very often subject to controversy.

12 Source: http://www.tableauxinteractifs.fr/le-tbi/de-linvention-du-tbi-a-aujourdhui/.

10

Modeling of Next Generation Digital Learning Environments

For Littré13 (1967, p. 1079), knowledge is used only in the singular and is defined as “understanding acquired by study, by experience”. According to the free encyclopedia Wikipedia, knowledge “designates an individual mental construct that can encompass several areas of knowledge”. In addition, according to the latter, understanding refers to a specific domain outside the subject: understanding of a language, of a discipline. For Catherine Chabrun, editor-in-chief of the Nouvel Éducateur14, who tries to bring some order to the definition of these terms, it is exactly the opposite: “Knowledge is a datum, a concept, a procedure or a method that exists at a given time outside of any familiar subject and which is generally codified in reference works […] Understanding is inseparable from a knowing subject. When a person internalizes knowledge, they transform that knowledge into understanding. They build this understanding. The same understanding built by another person will not quite be the same”. Insofar as the definitions proposed by Catherine Chabrun do not oppose those of Littré15, we willingly keep them because they correspond more to the perception that we have of their meaning. As for ability, it refers to the axis of development according to which the educated pupil or adult has to progress. Abilities are essentially transversal, they express a person’s potential (of a subject), independent of the specific contents of this or that disciplinary knowledge. They are not directly measurable. Take, for example, the ability to: – recognize: activity of description, of identification falling within classificatory logic and obviously of observation; it is also a question of knowing how to master different languages (mathematics, conventions for a circuit diagram, etc.); – explain: providing one or more explanation(s) by seeking to link causes and effects, by using theories, laws, principles and models, by transferring reasoning procedures already seen; – experience: testing, controlling, reassessing; critical thinking is implemented, as well as the need to verify through experience, reasoning or documentary research, to compare experimental procedures, technical processes, and group them; 13 Littré is a dictionary of the French language, abridged from the dictionary of Emile Littré of the French Academy by A. Beaujean, Editions de L’Erable, Paris, 1969. 14 Catherine Chabrun is the editor-in-chief of the Nouvel Éducateur, a pedagogical review of the Freinet movement. She tries to put some order in the definition of these terms on her blog of ICEM (Cooperative Institute of the Modern School – pedagogy Freinet), 25th November 2012, accessible at: http:// www.icem-pedagogie-freinet.org/node/3593. 15 Edition of 1967, p. 216 and p. 1079.

A System of Instrumented Activities

11

– realize: we appreciate here the quality of manipulation, the efficiency of the realization or the assembly, the correct use of the measuring devices, the “specific languages” (codes, mathematical tools) and the correction of the written expression. Let us remember that the important thing is not so much the definition of the words that we use, but rather our capacity to define these words before using them in an argumentative development, such as the one we suggest holding here. The key will also be to understand that as you learn, “you become someone else and you change your vision of the world and problems. Some do not realize it, others positively experience this intellectual and identifiable change, while others still show strong resistance” [PER 04c]. 1.1.4. Learning modes Let us return for a moment to the major modes of learning: “There are three main traditional modes of learning” [GIO 01, p. 95]. The first refers to the “transmissive” mode: we know, and we transmit to someone who does not know. This is the dominant practice (a conference situation). “Acquired by a ‘virgin’ brain and always available, the acquisition of knowledge is the result of a transmission” [GIO 01, p. 95]. It is an issuer that transmits data to a receiver. “In teaching, it’s the routine presentation of data, illustrated or not. At the museum, it is the exhibition of objects or documents accompanied by explanatory plates” [GIO 01, p. 95]. “The second tradition is based on training promoted to guiding principle […]. Learning is fostered by reward (‘positive reinforcement’ or ‘negative reinforcement’ punishments). Through such conditioning, the individual ends up adopting the appropriate behavior, one that avoids negative reinforcement” [GIO 01, p. 95]. The third is called the interactionist or construction model. “It starts from spontaneous desires and the natural interests of individuals […]. It highlights independent discovery or the importance of trial and error in the act of learning. The individual is no longer content to receive raw data; he/she selects and assimilates them” [GIO 01, p. 96]. 1.1.5. Learning modes in education To understand how technologies can affect learning, we must first clarify what is meant by learning. When we talk about the learning process, we are not referring to pedagogy or even didactics, even if all of these concepts are related. Cognitive

12

Modeling of Next Generation Digital Learning Environments

psychology is a science that seeks to answer an essential question: how do we learn? Pedagogy is, in a way, the art16 of teaching, in other words, it deals with methods and practices of teaching and education, as well as with, “all the qualities required to transmit any kind of knowledge”17. Didactics is something else, it is about a discipline. For example, we often hear about the didactics of mathematics. This mainly deals with the subject matter, its presentation, as well as the translation of “scholarly knowledge” into “knowledge to learn”. In the latter case, we speak of didactic transposition. As early as the 20th Century, pedagogical currents (or methods of teaching) were considered as coming from conceptions of education, which were themselves enriched by the development of research in cognitive psychology. The pedagogies would follow the following sequence:

Figure 1.1. From cognitive psychologies to pedagogies

It is from the different learning models (results in cognitive psychology) and following for each of them this specific sequence that we will show that so-called educational technologies do not affect the way we learn. On the contrary, each time a new model was set up for learning, the question was rather how technology could support the cognitive processes involved in each of these models: school radios and television for objectivist models, computer-assisted learning for behaviorist models, microworlds for the constructivist model, etc. 1.1.6. ICT, learning and pedagogical theories: an interrelated revolution According to this book, technologies would only reinforce processes already known and supposedly conducive to learning. Their “power” would only facilitate their realization, or even amplify certain already well-identified effects. It is 16 Way of doing a thing according to certain methods (definition of Littré, 1967, tome 1, p. 56). 17 Article “Pédagogie”, French Wikipedia, (2017, updated 26th January), available at: https://fr.wikipedia.org/wiki/P%C3%A9dagogie.

A System of Instrumented Activities

13

therefore by acting on certain procedural characteristics implemented during training that the technology would bring added value. For example, it could increase a student’s motivation or interest in a particular activity. Motivation and interest are already factors that are supposedly conducive to learning. Technology would amplify these results. Technology would act by a ripple effect that would resonate with already noted effects, and not necessarily in a direct way. For example, it is recognized that motivation promotes learning and that the use of certain technologies promotes motivation. It is only through “transitivity” that we believe that technologies promote learning. They obviously have other advantages and in particular undeniable pedagogical virtues. For example, they facilitate working independently or play a vital role in differentiated pedagogy by giving students the opportunity to evolve at their own pace. However, this is not the focus of our argument. We want to make it clear here that they do not affect the way we learn. Like the pedagogies that have followed the evolution of cognitive psychologies, technologies have been able to adapt to different models of learning without ever constraining them. Cognitive psychologies, pedagogies and technologies have therefore evolved simultaneously. This is at least what we have found by studying, for the purposes of research, these virtual learning environments. Georges-Louis Baron [BAR 11, p. 112] observed that, remarkably, “developments in computer-based learning environments have always operated in the more or less explicit context of theories related to learning and instruction”. To support this, we will cross-reference learning modes [GIO 01, p. 95] with the pedagogical models deriving from them, paying particular attention to the place occupied in these models by technology, all of course forming a system of instrumented activity that it has been possible to model. 1.1.7. From objectivist epistemology to school radio and television The 1960s and 1970s were dominated by a transmissive mode of learning, “strongly imbued with the theories of communication [SHA 75] which […] compared the media as a penetrative entity, an inert communicator disseminating information from the transmitter to the receiver” [DES 99]. It was assumed that the child knows nothing about the knowledge to be acquired. The role of the teacher is then to transmit the knowledge and that of the pupil is to listen attentively. 1.1.7.1. From objectivist epistemology to transmissive pedagogy This model is based on an “objectivist” conception of learning that dates back to the mid-1950s. Objectivism defends the existence of a knowledge of the world that exists outside the thinking subject. “The goal of the learner is to take ownership of it in order to replicate it; the goal of the teacher is to transmit it. Learning therefore

14

Modeling of Next Generation Digital Learning Environments

consists of assimilating this objective reality”18. It is therefore an epistemology according to which the learner’s activity is limited to assimilating the knowledge transmitted by experts. In this case, memory is described as “a repository in which knowledge accumulates” [DEP 09, p. 21]. The student is virtually passive during the teaching stage. From this mode of education came frontal instruction, in which the teacher, in front of the pupil, dispenses knowledge through lectures. The resulting pedagogical model is similar to the Shannonian model: a transmitter (the one who knows) transmits the data to a receiver (the one who does not know). 1.1.7.2. From transmissive pedagogy to school radio and television It was then that school radios and television made their appearance in education, responding perfectly to the teaching model of that time. For a long time, the Jesuit education was an example of this intellectualistic model, in which knowledge is simply transmitted. The transmissive mode of teaching existed well before school radios. For example, the Lasallian brothers were praising (since 1680) the merits of a transmissive approach to teaching to enable as many people as possible to attend school. Here, it was neither the school radio nor the school television that gave rise to this mode of teaching, but rather a widespread idea at the time that the human mind would be born as a “virgin” with regard to knowledge: a conception of teaching based on the tabula rasa model (blank slate). And thus radio and television transmission of knowledge was perfectly adapted to this mode of teaching. It also made it possible to “technologically” increase educational listening content on a large scale. It is in this sense that we say that technology helped to promote learning without changing the way of learning. It would have been very difficult indeed, if not impossible, to make “the voice of knowledge” heard on this scale without technology. 1.1.7.3. From transmissive pedagogy to multimedia With the arrival of multimedia, which combines text, still or moving images and/or sound, technology has prided itself on bringing real added value to learning. Dessus and Lemaire [DES 99, p. 1] tell us that from there came new theories: “These theories have been described as additive by Clark and Craig [CLA 91] (where more equals better), but they are not often empirically validated and rarely rely on a cognitive model of learning”. In the footnote, these authors state that Rogers and Scaife [ROG 95] identified “more equals better” theories which, for them, belong to folklore. Here are a few: – still images and diagrams are more effective than verbal representations; 18 Kozanitis [KOZ 05, p. 5], Bureau d’appui pédagogique, Polytechnique Montréal, September 2005, available at: http://www.polymtl.ca/bap/docs/documents/historique_ approche_enseignement.pdf.

A System of Instrumented Activities

15

– 3D representations are more efficient than 2D representations; – color images are more effective than black and white images; – hypermedia is more effective than sequential media; – multimedia is more efficient than single media; – virtual reality is more effective than multimedia [DES 99, p. 2]. 1.1.8. From behaviorism to computer-assisted learning Also called conditioning model, behaviorism is behavioralist. It is a psychological approach “which focuses on the study of observable behavior [of the subject] and the role of the environment as a determinant of behavior” [TAV 99, p. 182]. It is therefore interested only in the inputs and outputs, and not the cognitive system, which it considers as a “black box” [RAY 97]. The principle that justifies this choice is that it is not possible to know precisely the internal processes that act inside the brain. Behaviorism therefore sticks to the study of behaviors based on experimentation and observable and measurable things (stimulus and response) in order to understand learning. Behaviorism is often summarized by the simple schema in Figure 1.2.

Figure 1.2. Simple schema of the behavioralist approach

“The task of the psychologist [this applies to Skinner especially] then became that of studying the ‘reinforcement contingencies’, that is to say the relations between discriminative stimuli, responses and reinforcements” [DEL 99, p. 15]. Before discussing its characteristics, let us return for a moment to Pavlovian conditioning, which involves both external and internal stimulations19 that the organism is subject to.

19 “Generally speaking, learning consists of acquiring knowledge about the relationships between the events which constitute for the organism the stimulations (external and internal) to which it is subjected or that it generates its own responses. In this perspective, learning involves an internal representation of these ‘events’. […] if Pavlov’s dog salivates at the sound of the bell, it is because he ‘knows’ that this sound precedes the coming of food” [DEL 99, p. 15].

16

Modeling of Next Generation Digital Learning Environments

1.1.8.1. Pavlov’s conditioned reflexes (1849–1936) Physiologist Ivan Pavlov was awarded the Nobel Prize in 1904 for his research on the salivation reflex. He built an experimental device in which he tied dogs to a frame. Regularly, a bell would ring shortly before the distribution of dog food. After several experiments, Pavlov found that when the bell rang, the dog started to salivate. “These reflexes can be likened to an involuntary, non-innate reaction caused by an external sound signal. Pavlov developed the theory that the reactions acquired through learning and habit become reflexes when the brain makes the links between the sound signal and the action that follows”20. 1.1.8.2. Operant conditioning: Skinner (1904–1990) Pavlov’s theory was later adopted and developed by his followers. Among them, Watson [WAT 72a], influenced by the work Pavlov on conditioning in animals, formulated the psychological theory of stimulus–response (or classical conditioning). Skinner [SKI 95], who aligned himself closely to the Pavlovian theory, “proposed a psychology that does not refer to physiology or events described as ‘mental’, such as representations or expectations” [DEL 99, p. 18]. He also focused on the study of cause–effect relationships that could be established by observation by rising “against any introduction of intermediate variables, be they of a cognitive or motivational nature” [DEL 99, p. 15]. He therefore repeated Pavlov’s theory by applying it to learning with a particular focus: strengthening the response of the learner. It consists of giving a reward (and not a punishment) to the learning subject as soon as it has carried out the expected action. For example, if an animal activates a pedal, the food that will fall into its cage by the action of the pedal is a booster. Starting from this concept, Skinner then became interested in reinforcement contingencies. He sought to “show the possibility of accounting for the ‘occurrence’ of any behavior from the concepts of stimulus, response and reinforcement” [DEL 99, p. 15]. “This approach may seem unenthusiastic for psychology and the ‘Skinnerians’ sometimes appear as a chapel concentrating on variations on the same theme, published in a journal specific to this one. Their merit is often to offer simpler explanations (and that we must therefore prefer) than those of cognitivists” [DEL 99, p. 18].

20 “Classical conditioning”, Wikipedia (2015, updated 18th December), available at: https://en.wikipedia.org/wiki/Classical_conditioning.

A System of Instrumented Activities

17

Skinner also introduced the concept of maintaining satisfaction (through reward) and promoted a positive learning process: stimulus, response, reward (when a correct answer is given) and reinforcement. The reinforcement works by reaffirming the thing learned by repetition of a given correct answer. “From the point of view of teaching, behaviorism therefore considers learning as a lasting change in the behavior resulting from particular training. It assumes that the acquisition of knowledge takes place in successive stages. The transition from one level of knowledge to another is achieved by positive reinforcement of the expected responses and behaviors” [CHE 15a, CHE 15b]. 1.1.8.3. From behaviorism to programmed instruction Following Skinner’s works, a new method of teaching called programmed instruction (or linear curriculum) was born. The following type of programmed lesson, devised by Skinner himself, is the most orthodox adaptation of the principles or ideas discussed so far [POC 71, p. 55]. It allows the individualization of learning through a breakdown of content. This content is given in micro-steps, which are easy to work through. Here is an example (Figure 1.3) of learning to spell the word “aerodynamic” correctly.

Figure 1.3. Sample sheet for learning to write the word “aerodynamic” [POC 71, p. 55]

18

Modeling of Next Generation Digital Learning Environments

1.1.8.4. From programmed instruction to teaching machines The pedagogy that can be compared to this model is the one introduced in teaching–learning situations, self-correcting exercises that automatically process learner responses and respond with the answer or comment until the correct answer is provided. This processing is usually supported by automatons. The first of these was probably the teaching machine proposed by Skinner himself. The hypothesis is that knowledge can be taught by a teacher or a machine, since it is a pre-programmed content that we want to transmit. Concretely, these machines (or automata) make it possible to revise or to deepen certain notions. The device proposed by Skinner: “does not have a keyboard, the student writes his answer on a roll of paper through the window designed for this purpose. Once the answer has been entered, pressing the lever will drag the paper roll and drag the answer underneath a transparent cover. At the same time, in the window used to present the learning situation, the exact answer appears so that the student can evaluate their answer by comparing it to the correct answer provided”21. 1.1.8.5. Computer-assisted learning Computer-assisted learning replaced Skinner’s teaching machine using computers and specific software: tutorials and exercises (a series of training exercises). The computer and its software bring: – a closer connection between the information given and the verification of understanding by the exercise; – more precise and immediate information in case of error; – progression more adapted to the pace of the learner. The tutorials and exercises therefore appear as artifacts perfectly adapted to the behavioral principles of learning by playing on their ability to significantly reduce the time between the response given by the learner and reinforcement delivered in return by the teacher. These exercises and, more generally, the computer-assisted

21 “Introduction à la technologie educative”: syllabus written by Christian Depover (2001, part 2) as part of the Unité de technologie de l’éducation (UTÉ) which is a service of the University of Mons-Hainaut (UMH) in the field of educational technology, TECFA website of the Faculty of Psychology and Education Sciences (FPSE) of the University of Geneva, available at: http://tecfa.unige.ch/tecfa/teaching/staf11/0304/basespsychopeda.pdf.

A System of Instrumented Activities

19

learning devices, thus make it possible to acquire and reinforce certain automatisms for learners by playing a supportive and in-depth role. Here again, technology supports and reinforces well-established methods of learning without affecting the epistemologies which they derive from. 1.1.9. From construstivism to microworlds This model comes from Jean Piaget’s research, who, contrary to the previous model, considers that we can study what happens in the brain. Piaget believes that knowledge is built: our brain develops at the same time as our body and we assimilate knowledge from experiencing and discovering the environment: “Constructivists believe that each learner constructs reality, or at least interprets it, based on their perception of past experiences. According to them, knowledge is not a reflection of reality as it is, but a construction of it. That said, constructivists do not reject the existence of the real world. They recognize that reality imposes certain constraints on concepts, but maintain that our knowledge of the world is based on human representations of our experience of the world” [KOZ 05, p. 5]. Thus, Piaget’s theory is based on the idea that we build knowledge from the reconceptualization of experiences. The Piagetian School of Geneva borrows the concepts of assimilation and accommodation from Darwin’s theory of evolution [GIO 01, p. 97] to explain the process driving self-construction works. “Every organism assimilates what it takes from the outside to its own structures, including information retrieved by its perceptions” [GIO 01, p. 98]. It is the progressive integration of new objects into existing patterns that constitutes the process of assimilation. For Piaget [PIA 69, p. 45]: “Knowledge derives from action, not in the sense of simple associative responses [in reference to behaviorism], but in a much deeper sense which is that of the assimilation of the real to the necessary and general co-ordinations of the action. To know an object is to act on it and transform it, to grasp the mechanisms of this transformation in connection with the transformative actions themselves. To know is thus to assimilate the real to structures of transformation, and these are the structures that intelligence elaborates as a direct extension of action”

20

Modeling of Next Generation Digital Learning Environments

“This process is accompanied in return by an accommodation, that is to say, a modification of the organs on the biological level or intellectual instruments on the cognitive level. If the subject wants to assimilate knowledge, he/she must be able to permanently adapt his/her way of thinking to the situation’s requirements” [GIO 01, p. 98]. The image of the tectonic plates that move on the surface of the Earth is a metaphor quite similar to the way Piaget envisages learning: our brain consists of schemas (plates) that are jostled and disturbed by the assimilation of new knowledge and stabilized by integrating this new data into our system of understanding. “A schema is the structure or organization of actions as they transform or generalize when repeating this action under comparable or similar circumstances” [PIA 04, p. 11]. Therefore, learning is essentially achieved through action and this theory finds its application in pedagogy by offering activities for learners that lead them to think, to develop their critical thinking and thus to change the schemas of their learning system. 1.1.9.1. From constructivism to active pedagogies (simple self-development) These so-called “active” methods used in schools and a certain number of places of investigation are built on this constructivist model of learning. They are based on students’ needs and initiatives. According to Cousinet, “active methods are not teaching but learning tools, which must be put exclusively in the hands of students, any teacher introducing them into his or her class will be prepared not to use them, and in so doing will refrain from teaching”22. The lecture is therefore replaced with group work and the teacher’s status changes. The teacher changes from the diffuser of knowledge to the constructor of learning situations; situations in which students will make assumptions as soon as they are confronted with new problematic situations. They will then propose an experimental protocol to validate or invalidate the hypotheses. “New education” brings together all the innovative educators who, with very different techniques, have experimented with these methods: Montessori, Ferrière, Cousinet, Freinet, Dewey, Kerschensteiner, etc. Following the intuitive methods of Pestalozzi and Fröbel, these so-called “active” methods lead the teacher to create living activities whose actions are in keeping with reality. Emphasis is placed “on the active aspect of teaching, on the teacher’s efforts to ensure that the child acquires skills, knowledge and indispensable notions by him or herself, by following his or her curiosities, by freely developing his or her activities” [LEI 70, p. 307].

22 Dictionnaire encyclopédique de l’éducation et de la formation, 3rd version, published by Philippe Champy and Christiane Étévé, Éditions Retz/S.E.J.E.R, 2005, p. 228.

A System of Instrumented Activities

21

1.1.9.2. Active pedagogies to hypermedia, microworlds and school newpapers The teacher who practices these pedagogical methods must therefore offer the learner an environment. In this environment, the learner can explore and discover abstract domains for themselves through their personal activities. Seymour Papert, a long-time contributor of Piaget’s, was one of the first to wonder how the computer could take part in this constructivist approach. For him, the time for programmed learning from behaviorism was over. The child had to be at the center of their own learning, an actor in the construction of their own knowledge. He remarks on this subject: “In many schools today, computer-assisted learning means that the computer is programmed to teach the child. We could say that the computer is used to program the child. In my vision of things, the child programs the computer and, in so doing, gains control of one of the most modern and powerful technological elements, while establishing intimate contact with some of the most profound notions of science, mathematics, and the art of building intellectual models” [PAP 81, p. 15]. Specifically, the computer environments that allow this meeting are hypermedia that offer the learner a space of exploration and microworlds that provide them with the means to discover abstract areas. In this context, Papert seeks to provide students with intelligent tools, such as a programming language, to introduce them to abstract domains. For him, “all the conceptual tools necessary for the development of a scientific field must be organized”. Papert calls this the microworld. “[Papert adds] that children have to have an object with specific properties, in terms of movement, for example, around which they will gradually develop a culture in a given field. […] The object that helps children is for example a small arrow that moves on a computer screen. From there Papert’s famous Logo language (with turtle graphics) was born” [PER 83, p. 94]. Logo was initially intended to teach programming to children from the ages of 10 to 12: it is the child who programs the computer and not the opposite, as Papert wished. The most popular microworld successor to Logo (which has been declining since its creation in 1972) was the Squeak environment, which is a dynamic programming environment, written in the Smalltalk language. Then, software such as Cabri Geometry for primary school students or GeoGebra for high school students succeeded Squeak. These environments are probably the most current microworlds of the French education system. These are virtual learning

22

Modeling of Next Generation Digital Learning Environments

environments for geometry where students can easily construct figures using all kinds of concepts such as perpendicularity, parallelism, tangency and so on. “[…] the software promotes the implementation of analytical procedures in the reproduction or construction of geometric figures […] The obligation to communicate to the software a geometrical method of construction thus makes it possible to characterize the geometrical object …” [BEL 92]. Once again, the aim was not to find new applications in the education world imperatively, but rather to start from existing pedagogical approaches (here, active pedagogies) and to focus on how these technologies could usefully participate in these approaches. As Perriault says: “Papert does not start from the use of the machine to achieve pedagogical effects, but, in the opposite direction, he proposes a theory of the acquisition of knowledge and this leads to a use of the machine that is consistent with it” [PER 83, p. 94]. The purpose of these tools was to provide children with “a large, easily accessible space of freedom where they can express their ideas and explore the consequences”23. Students have the opportunity to express their own representations of perceived phenomena while having the opportunity to test them in digital environments built for this purpose. In return, these environments send back information that matches their initial representations while acting on them. New models of representations are thus constructed through this iterative process which is nothing other than a learning process24. From the computer-tool, we pass to the computer-mediator, which allows the students to express their representations of the perceived phenomena, and to act in return on these models of representations, by possibly making them evolve towards a state of higher knowledge. Many other “technologies” have also demonstrated their ability to adapt to active pedagogies. Examples include students who produce and publish school newspapers using DTP (desktop publishing). This activity falls well within the so-called “active” pedagogy and in particular the “project-based” pedagogy so dear to Dewey. “Project-based learning [according to this definition] involves elaborating according to a plan, a study of the conditions, and the intellectual anticipation of the consequences” [LEI 70, p. 134]. The computer is not at the forefront of the educational project, it only supports it by facilitating the student’s task. Long before 23 “Micromonde”, French Wikipedia (2016, updated 6th January), available at: http://fr.wikipedia.org/wiki/Micromonde. 24 If we stick to the definition given by Grégory Bateson [BAT 77, p. 303]. For him, “the word training undoubtedly indicates a change of one kind or another”.

A System of Instrumented Activities

23

the use of computers in schools, Freinet had already invited his students to use the printing press to publish the texts they wrote. The children were then intellectually stimulated: “[with] the prospect of seeing their text printed in ‘the school newspaper’, word spread among the fans of the movement, and even among inter-school exchanges […] Thus arose tales, stories and children’s poems illustrated by children, which Freinet was proud to publish in small brochures called ‘Enfantines’ and which can serve as a reading book” [LEI 70, p. 311]. The computer merely replaced the printing press originally introduced into the classroom by Freinet. 1.1.10. From social constructivism to CLE, CSCL and cMOOC A contemporary of Piaget, psychologist Lev Vygotsky developed an interactionist theory that makes learning dependent on social interactions. It is indeed in social environments that we find “the cultural traces of the intellectual and technological conquests achieved by men during their history” [CRA 01, p. 145]. For Vygotsky: “The acquisition of higher mental functions depends both on the social exchanges that the individual can have through language and on the transmission, from generation to generation, of the cultural heritage that characterizes a given society at a given moment in its history”25. Without doubt, it is for this reason that this theory has been described by Vygotsky as historico-cultural (of psychism). It is from this theory that new learning methods have been developed, such as “social constructivism” for example. This theory is subsequent to Piagetian constructivism and the historico-cultural theory of Vygotsky. It is an extension of the latter and aims to show that social interactions between peers help to influence cognitive development. We owe [DOI 81; DOI 97] this introduction of the social element to Piagetian constructivism to Doise and Mugny in particular as well as the resulting works. Socio-constructivism thus introduces a new dimension: we learn better through contact with others and by exchanging our experiences with them. A phrase traditionally associated with Vygotsky is that “what a child can do today with

25 [DEP 01, part 5], /basespsychopeda.pdf.

available

at:

http://tecfa.unige.ch/tecfa/teaching/staf11/0304

24

Modeling of Next Generation Digital Learning Environments

assistance, she will be able to do by herself tomorrow” [VYG 97, p. 355]. He therefore introduces mediation, that is, dialogue with another (another learner, expert, teacher, etc.) as a determining parameter in the learning process. 1.1.10.1. From social constructivism to active and collaborative pedagogies From a pedagogical point of view, Vygotsky has the same desire to support children towards autonomy through action, but the emphasis this time is on “doing together”, with particular emphasis on collaborative activities. According to this method, the role of the teacher is to favor situations in which the child is an actor of their/them learning, and also in which their/they interact socially with their peers. These situations should encourage them to verbalize, to explain how they go about doing one thing and to compare it to the strategies of others. “Learning with others opens up a whole series of development processes for the child which occur only in the context of communication and collaboration with adults or peers, but which will become the child’s own achievement afterwards” [VER 00, p. 23]. Vygotsky therefore proposes a third model of the relationship between development and learning. Contrary to Piaget, he recognizes that there is logical development to respect, but considers that learning must anticipate development within an area that he names the zone of proximal development. He defines it as “the difference between what a learner can do without help and what the learner can do with help” [VER 00, p. 22] further adding that: “Each psychological function in the child’s cultural development appears twice: first, on the social level, and later, on the individual level: first, between people (interpsychological) and then inside the child (intrapsychological)” [SCH 85, p. 111]. Thus, Piaget proposes a biological method of learning, moving from cognitive to social development, whereas Vygotsky proposes a social method of learning, moving from social to cognitive development [AST 01, p. 108]. In addition, since peer-to-peer problem-solving interactions play a crucial role in the construction of knowledge according to this Vygotskian theory, a teacher must not only promote group work and communication, but also recognize the importance of facilitating relationships and mediation.

A System of Instrumented Activities

25

1.1.10.2. Active and collaborative pedagogies to CLE, CSCL and cMOOCs The assumption that knowledge is acquired and socially produced has profoundly changed the way virtual learning environments (VLEs) for long-distance learning are designed and managed. In its early days (i.e. in the 1990s), long-distance education adopted teaching methods based on the transmissive model and without any possibility of peer-to-peer communication. Its operation was modeled, at least in France, on the epistolary model of the National Center for Distance Education, with the only difference that the lessons’ contents were channeled electronically rather than by post. Digital networks were then used only to facilitate their transmission and reduce the time needed for homework. This resulted in an increase in the number of “dropouts” compared to what can be recorded in formal education, in other words, face-to-face. At this time, Jacquinot [JAC 93, p. 57] already considered that the concept of distance had diversified over time. This distance could be “at the same time spatial, temporal, technological, but also psychosocial …”. With this in mind, distance learning platforms have since been designed and constructed based on the socio-constructivist model to recreate the social link that distance has previously broken. The idea of socio-constructivism applied to distance learning was to provide learners with relevant information and a communication space (synchronous and asynchronous) to facilitate collaboration around problemsolving situations. According to Gilly [GIL 95], “the possibilities of interpersonal interaction at a distance make it possible to highlight socio-cognitive conflicts that are supposed to facilitate the acquisition of knowledge”. More generally: “The values associated with constructivism and the principles of learning from the cognitivist current lay the foundations for collaborative learning, namely the recognition of the multiplicity of representations and knowledge, which find their viability through social and cultural anchoring; interaction between peers and with the trainer; negotiation and validation of knowledge; cognitive and social reflexivity; recognition of the individual dimension of the learning process; and the precedence to be given to the coordination of collective processes” [HEN 01, p. 23]. “Over time, the production of knowledge is gradually shifting from traditional disciplines to a distributed appropriation of knowledge in new social contexts. New, decentralized, deterritorialized forms of organization emerge and accompany these changes. Intranets, extranets, virtual communities of cooperative development tend to

26

Modeling of Next Generation Digital Learning Environments

remain through vast memory networks whose knowledge is made universally accessible in new dynamic configurations” [HAR 01, p. 28]. Thus, in 1995, we saw the development of “new forms” of distance learning, without centralized control: the trainer becomes a guide within a group of learners, whereas in the past, the trainer was in charge of conveying knowledge. These new methods are based on the conjunction of three factors: – the benefits that can be drawn from the learning theories we have just mentioned; – the possibilities offered by the Internet; – the new need for complex problem-solving in organizations. They therefore jointly use the tools of collaborative and cooperative work and distributed technologies. These tend to reduce the concept of space by cognitively bringing together the various actors involved. Built around a multi-agent system, they offer the possibility of distributing knowledge and apply to network situations, which offer solutions in terms of sharing, collaboration and cooperation. These technologies make it possible, in the realization of a common motivated goal (around a project, a task to be accomplished), to integrate expertise, skills and individual knowledge within a knowledge base. The basis thus constructed leads to comments, reactions and adjustments that enrich it again. In this process, we see a retroactive effect, where singular knowledge nourishes the knowledge base. This, which embodies collective knowledge, in turn promotes individual learning. This process of dynamic knowledge appropriation is commonly referred to as distributed learning. It is based on the concept of distributed cognition which “is interested in the structure of knowledge (representations) and their transformation. It differs from the traditional models of cognitive science in that the object of study is no longer solely in the minds of subjects but includes the processes of cooperation and collaboration between subjects. The unit of analysis […] is a cognitive system composed of the individuals and artifacts they use” [JER 96]. In this situation, the learner decides on their own path and their own actions and becomes an actor in the construction of their knowledge (constructivism). The learner can have access to different cognitive tools located in physically remote systems. More specifically, these tools include those mentioned previously (hypermedia, for example) and those specific to network collaborative work to

A System of Instrumented Activities

27

produce and distribute knowledge. These tools exist concretely on computer platforms coined computer-supported collaborative work (CSCW), also known as groupware, on so-called collaborative work platforms belonging to the computer-supported collaborating learning (CSCL) domain or in constructivist learning environment (CLEs). More recently, (in 2008 for CCK0826), MOOCs joined this family of distance learning devices. All of these virtual learning environments (VLEs) bring about a profound change in the way we work, but in no way alter the way in which we learn. The student is characterized by their intermediary status in a virtual group in which they interact, cooperate, exchange information synchronously or asynchronously and make decisions in an electronically constructed sphere. Technology facilitates the acquisition of knowledge through socio-cognitive conflicts that make it possible without causing them. 1.1.11. Learning in an open network: from connectivism to MOOCs 1.1.11.1. Connectivism We could not end this chapter without mentioning the so-called socio-informatic theory [CRI 12] or “neosocioconstructivist” [GUI 04] approach, which is better known as connectivism. This theory of learning, developed by George Siemens [SIE 05], one of the founders of CCK08, is based on a critical analysis of the models studied previously to propose a unifying theory of networked technology using interposed technologies. Siemens is convinced that “these theories were developed at a time when learning had not yet been touched by technology” [SIE 05, p. 1]. From our point of view, expressing it in this way shows a lack of knowledge of the role that these technologies played well before connectivism came into being. This is reflected in the writings of Seymour Papert, who was already wondering what place the computer could have in a constructivist approach. Indeed, since the 1960s, Papert had already collaborated with Piaget before developing the Logo programming language at MIT. It must be recognized, however, that in the last 20 years, technology has dramatically changed the way we live and communicate and continues to do so. Does this mean, as Siemens argues, that it has “reorganized the way we learn” [SIE 05, p. 1]? According to the theory that we have supported since the beginning of this chapter, it is less certain. Let us highlight again that the polysemous nature of the word “to learn” can lead to very different interpretations. For Siemens, learning refers mainly to a set of means available to the learner for learning. Based on the “half-life of knowledge” concept developed by Gonzalez [GON 04] which represents “the time span from when knowledge is gained to when

26 CCK08 (The Connectivism and Connective Knowledge course) is considered the first MOOC (massive open online course).

28

Modeling of Next Generation Digital Learning Environments

it becomes obsolete it proposes to fight against the shrinking of the half-life of knowledge by developing new instruction methods”. This at least has the advantage of confirming that Siemens rather considers connectivism as a means of learning faster than as a cognitive process per se. After recalling the criticisms of this theory, Cristol [CRI 12] pointed out that “connectivism, by integrating already existing theories, is more of an educational proposal than a theory”. According to this author, “the implementation of MOOCs is a practical example of connectivist pedagogy […]”. Participants self-teach and self-animate in a lively space. Connectivism thus emphasizes informal learning, which is gradually taking precedence over formal learning. According to George Siemens: “Formal education no longer comprises the majority of our learning. Learning now occurs in a variety of ways – through communities of practice, personal networks, and through completion of work-related tasks” [SIE 05]. Several recent experiments aim to show that learning can occur by simply being in contact with technology without the help of a teacher. Sugata Mitra, for example, has conducted a ten-year study that “demonstrates that children are able to learn (co-learn) alone, without a teacher, that is, by themselves with a computer, Internet or course materials. He concludes that education is a self-organizing system where learning is an ‘emerging phenomenon’”27. Siemens suggests replacing the behaviorist, constructivist, and even socioconstructivist theories with a theory that seeks not only to study how learning occurs “inside the person” but also outside. In other words, he seeks to construct a model that rejects the idea that learning is only the result of an individual and internal activity: this learning would also be a “role of the environment and the communication tools available” [GUI 04]. “A central tenet of most learning theories is that learning occurs inside a person. Even social constructivist views, which hold that learning is a socially enacted process, promotes the principality of the individual (and her/his physical presence – i.e. brain-based) in learning. These theories do not address learning that occurs outside of people (i.e. learning that is stored and manipulated by technology)” [SIE 05].

27 “Connectivism”, Wikipedia (2015, https://en.wikipedia.org/wiki/Connectivism.

updated

3rd

August),

available

at:

A System of Instrumented Activities

29

In addition, according to Siemens [SIE 05], learner-centered theories also fail to describe how learning occurs within organizations. For example, the ability to assess the relevance of learning one thing over another is a meta-competence that is applied even before the act of learning begins. In fact, the study of internal learning processes is essential when the knowledge that is presented to the learner is scarce, selected by a teacher, and the question does not arise whether it is relevant or not relevant to acquire this knowledge. However, in the opposite case, meaning when the knowledge is abundant and varied, it is essential to be able to quickly assess the quality and the relevance of the knowledge considered before even starting to learn it. In other words, the ability to synthesize and recognize links and patterns (recurrences, patterns) is a valuable skill. When learning theories are seen through technology, many other important questions arise. For Siemens [SIE 05], the natural temptation for theorists is to continue to revise and evolve learning theories as soon as initial conditions change. For him, this is possible only to a certain extent. Because when the initial conditions are changed too much, the adjustment of the models to these variations becomes impossible. A completely new approach becomes necessary and the questions to be explored must be dealt with in other contexts: chaos and network theories, for example. Let us take a closer look at these two aspects. Contrary to constructivism, which asserts that learners are likely to learn better by doing meaningful tasks, chaos theory asserts that (pre) sense exists; the challenge for the learner is to recognize the patterns (recurrences, models) that are otherwise hidden. “Unlike constructivism, which states that learners attempt to foster understanding by meaning making tasks, chaos states that the meaning exists – the learner’s challenge is to recognize the patterns which appear to be hidden” [SIE 05]. Using the analogy of the “butterfly effect”28 by Edward Lorentz [GLE 87], George Siemens recalls that sensitive dependence on initial conditions profoundly impacts what we learn and how we act based on our learning. Decision-making is indicative of this. If the underlying conditions used to make decisions change, the decision itself is no longer as correct as it was at the time it was made. Connectivism is built on the idea that the decisions we make are based on rapidly altering

28 “Does the flap of a butterfly’s wings in Brazil set off a tornado in Texas?”.

30

Modeling of Next Generation Digital Learning Environments

foundations. The ability to recognize and adjust to pattern shifts is a key learning task. The network can be simply defined as a set of connections between entities. “Computer networks, power grids and social networks all function on the simple principle that people, groups, systems, nodes, entities can be connected to create an integrated whole. Alterations within the network have ripple effects on the whole. […] Personal knowledge is composed of a network, which feeds into organizations and institutions, which in turn feed back into the network, and then continue to provide learning to [the] individual. This cycle of knowledge development (personal to network to organization) allows learners to remain current in their field through the connections they have formed” [SIE 05]. For Siemens: “Connectivism is the integration of principles explored by chaos, network, and complexity and self-organization theories. Learning is a process that occurs within nebulous environments of shifting core elements – not entirely under the control of the individual” [SIE 05]. In addition: “Connectivism is part of several phenomena specific to current professional activities: chaos (everything is connected), complexity, networks and self-organization. What constitutes the heart of connectivism theory is the role of links and flows between individuals and computers that accelerate them and not just the content of knowledge” [CRI 12]. The founding principles of connectivism outlined by George Siemens are as follows: – learning and knowledge rests in diversity of opinions; – learning is a process of connecting specialized nodes or information sources; – learning may reside in non-human appliances;

A System of Instrumented Activities

31

– the capacity to know more is more critical than what is currently known; – nurturing and maintaining connections is needed to facilitate continual learning; – the ability to see connections between fields, ideas, and concepts is a core skill; – currency (accurate, up-to-date knowledge) is the intent of all connectivist learning activities; – decision-making is itself a learning process. Choosing what to learn and the meaning of incoming information is seen through the lens of a shifting reality. While there is a right answer now, it may be wrong tomorrow due to alterations in the information climate affecting the decision. It is therefore necessary to convince oneself that: “The only relevant information to produce transformations in an organization is not in one head but in many. Incidentally connectivism advocates for teams composed of individuals with different points of view and who are able to receive and address criticism” [CRI 12]. According to connectivist theory, what needs to be developed to learn in our knowledge economy is the ability of individuals to form connections between the many sources of information in order to create models of useful information29. 1.1.11.2. From connectivism to MOOCs As we have previously seen, connectivist theory was originally taught in a traditional manner at the University of Manitoba by George Siemens himself and Stephen Downes30. Called Connectivism and Connective Knowledge, their course was later opened to all Internet users and “2,300 participants took part online free of charge in addition to the 25 students enrolled. They were free to participate and enrich the course with the tools of their choice” [CIS 13, p. 3].

29 Interview of George Siemens by Jacques Fayet and Matthieu Cisel (FFFOD) as part of the e-Learning days in June 2013, available at: http://cursus.edu/institutions-formationsressources/formation/20587/rencontre-georges-siemens-nous-parleconnectivisme/#Vp9LQ9TSk1N. 30 Stephen Downes was then a member of the National Research Council on Connectivism.

32

Modeling of Next Generation Digital Learning Environments

It was the decision of its co-founders to put this course online in June 2008 to further test the effectiveness of this theory, which marked the starting point of MOOCs, hence its first name: CCK08 (08 for the eighth version of the course)31. The principle was to focus, during the different connections, on the opportunity to create links between knowledge in order to create new ones. As a matter of fact, Stephen Downes’ first post discussion of the CCK08 course indicates that the first week of the course would begin by mapping some of the different areas of discussion and it would be necessary to begin by sorting out the differences between knowledge about networks and knowledge created by networks. For Stephen Downes, the connectivist theory of learning “is about making meaningful connections”. In this course, the content was not built or brought by the teacher but developed with the students themselves. The latter were therefore led to participate in and enrich the course. For this, they were free to do research to make their own contribution to the course. This serves as a reminder that: “Course content is not in one place but may be located anywhere on the Web. The course therefore consists of sets of connections linking the content together into a single network. Here is the structure of the CCK08 course network [see Figure 1.4]” [SIE 11]. The two co-founders of CCK08 specify that: “the idea behind the title of this course is important because it stems from the theory of connectivism that says learning in the digital age is based on building connections and relevant networks. This idea of connecting with each other to build knowledge is one of the main dynamics of a MOOC. Course participants were encouraged to develop their own online presence to add distributed resources to this network. Course authors then used a content aggregation tool to bring all the content into one place” [SIE 11].

31 Available at: http://connect.downes.ca/archive/08/09_15_thedaily.htm.

Figure 1.4. Course network structure CCK08. Redrawn from [SIE 11]. For a color version of this figure, see www.iste.co.uk/trestini/learning.zip

A System of Instrumented Activities 33

34

Modeling of Next Generation Digital Learning Environments

Remember that “in line with this experience, many MOOCs will be set up in different institutions” [CIS 12, p. 3]. This included the first Udacity course created by Sebastian Thrun in August 2012, which fundamentally differed from the previous one (CCK08) in the pedagogical approach it offers. Moving away from the constructivist model of learning, the Udacity course is based on a much more transmissive approach to learning than that of its predecessor. The video clips at the very beginning of each class feature renowned teachers who teach in the same way that they would do in a lecture theater. The course is characterized by its transmissive, masterly and frontal aspects. As a result of these two experiences, many MOOCs will be created (and continue to be created) around the world, reproducing, for the most part, one or other of these two pedagogical models. These two examples have been chosen not only because they have been recognized as the first in a long series, but also because they illustrate this difference in the pedagogical approaches of the MOOCs that will follow. 1.1.11.3. Contrasts between pedagogical models associated with MOOCs While CCK08 is based on a connectivist approach to learning, itself based on an established model which is described in detail in the previous section, its successor Udacity relies on a more traditional approach to learning. Often supported by prestigious universities, the Udacity platform offers courses led by their most renowned professors. Also unlike CCK08, these courses are very structured. Most of them (in the form of video modules) reproduce the model of frontal and transmissive teaching found in university lecture theaters. Registered users have relatively little influence on the content. It is the same for almost all of the MOOCs that followed, like Coursera or edX32. The first MOOC (CCK08) is commonly part of what is known as the cMOOC family. Later MOOCs like Udacity, Coursera and edX are part of the xMOOC family. In xMOOCs, the teacher(s) build a structured course that they intend to teach to their students. This is the classic classroom model, except that, thanks to the Internet, the course is potentially aimed at thousands of students. This remains a very top-down approach to teaching: the teacher knows and divulges his knowledge. The “x” in xMOOCs comes from Stephen Downes, who wanted to differentiate MOOCs as he conceived them (i.e. connectivist) from other, non-connectivitist, MOOCs. He noted that “the origin of ‘x’ and the use of ‘x’ in applications such as ‘TEDx’ or ‘MITX’ were intended to indicate that these programs were not part of a

32 This will not always be the case, for example the MOOC ITyPA is more inspired by CCK08.

A System of Instrumented Activities

35

basic offering but that they were somehow extensions”33. It was also used in some MOOCs in the United States as for example in “edX”. He began to call the MOOCs from Coursera, Udacity and edX “xMOOCs”. It was only later that he called the others “cMOOCs”. Downes contends that xMOOCs does not mean “EXtended MOOCs”, but this term refers rather to a MOOC perceived as an extension of something else. Connectivist MOOCs are called cMOOC, the “c” obviously referring to connectivist. More localized initiatives like the MOOC ITyPA are also very cMOOC-oriented. If we observe the content of a cMOOC, we see that in the end, there is not really any content: the participants progressively communicate what they know to others and add hyperlinks to other content that they find useful. In the spirit of these courses, the teacher in charge of them does not have the role of transmitting knowledge but of facilitating interactions between participants. There is no proper creation of educational resources; the resources are found on the Web and are aggregated by appropriate techniques. In a cMOOC, each registrant can potentially become a “teacher” who makes a difference in the course. When there are no teachers at all (which can happen), there are nevertheless organizers who structure the discussions and the proposed content. 1.1.11.4. MOOCs that revive the flipped classroom concept (flipped school) As we have seen above, the traditional model of university teaching is based on a series of lectures followed by tutorials. The lecture theater is usually the place where knowledge is transmitted to students by the teacher. The latter also takes the time to answer the questions asked by the students and returns to the difficult or poorly understood parts when necessary. Most often, the course is followed by tutorials during which training exercises are offered. The emergence of online video courses, such as those offered by the Khan Academy or those present in MOOCs, makes university knowledge much more easily accessible online, provided however that the discipline studied is not too uncommon. As Michel Serres [SER 12] says, knowledge is now “at the end of our thumbs”. These new conceptions of education are shaking up current forms and seriously questioning them. Awareness of a radical change in the way of teaching is now a reality, both for teachers and students. For the teacher, one of the problems will be to continue to interest their students if the course they offer is already online. How, for example, can the courses of distinguished professors of international renown be acquired? It is also worth mentioning the online courses judged by learners as “easier to understand” or “better presented” than those of the teacher they have in class. How should the teacher react to a student who has attained a high level of scholarly knowledge well before class? 33 Submitted by Stephen Downes himself and shared on 9th April 2013 on Google+, available at: https://plus.google.com/+StephenDownes/posts/LEwaKxL2MaM.

36

Modeling of Next Generation Digital Learning Environments

This case is far from exceptional. On some MOOCs, it is now easy to find very popular courses that students take for weeks before they even begin their traditional training. How can we then ensure suitable educational progression? For Marcel Lebrun: “The fact that most of the resources are available on the Internet is changing the roles of the teacher and the student, between whom the boundaries fade: by gathering information around him/her, in society, the student acquires knowledge that the teacher does not necessarily have, which places the teacher in a learning position” [BLI 14]. These singular practices are not necessarily seen in a bad light by teaching staff, and some even see the benefits. Imagine that all students arriving in class have already acquired online knowledge of the subject that the teacher has to deal with. They would only have to strive to raise the level of students by offering them more difficult problems, which would result in a significant cognitive gain. On the student side, many of them think that digital technology opens up new possibilities that must be exploited as soon as possible at university. If the knowledge to be acquired is easily accessible, why it is necessary to attend a class? University students are not the only ones to ask this question. In primary schools and high schools, students continue to advocate Internet use for learning. Many high school and college students complain that they are forced to listen to their teachers’ classes for countless days when they can listen to them more freely online. The course can be interrupted and resumed one to two hours later, at their convenience. (As for remaining questions, the school, the university or the teacher would obviously always be present to answer them …) We thus find the principle of a flipped classroom: the teacher produces a course that they make available to pupils or students (often electronically) before the meeting takes place. These may be video clips or more traditional online resources (text, images). Students watch and listen to them at home, on the bus, on the Internet, on DVDs and on their smartphones depending on the technologies they have or can use. Back in class, the students comment on the notions and concepts they have viewed and ask for clarification from their teacher. Diagnostic evaluations are sometimes planned before even debating them (tests, projects, essays, etc.). The time of the meeting is when the students pose questions, discussing recently acquired knowledge with their teacher or knowledge in the process of acquisition. The traditional course is thus replaced by a dialogue course. For Dufour:

A System of Instrumented Activities

37

“The flipped classroom approach takes place at home independently, with low cognitive, level activities to promote collaborative in-class work and high-level cognitive learning tasks, encouraging activity and the cooperation of students” [DUF 14]. The flipped classroom is now a huge trend [ROB 12], both in higher education and in primary and secondary education34. This concept, widely practiced in the United States and Canada, is developing in France and is already found in all levels of education. Xavier de La Porte tells an interesting story about this in LeMonde.fr (25th October 2013): “I’m going to tell you about an experience recently told in the New York Times by Tina Rosenberg, which is the starting point of a movement that bears the pretty name of ‘inverted school’ or flipped school. It all started a few years ago in Clintondale, north of Detroit, in a far from privileged region. The principal of Clintondale School had been recording videos on baseball techniques and posting them on YouTube for his son’s team. He realized that not only did the young players watch these videos, but they watched them several times … They assimilated strategies and it left more time for hands-on work at practices. He decided to apply this to one of his classes (the weakest) and the results were spectacular. He then applied it to the whole of his school, which also produced spectacular results”35. The first experiments conducted in the English-speaking world and in universities were carried out by Eric Mazur. At that time, he began research on new forms of teaching based on collaborative work. This approach encourages the child to be active with their peers during learning. He published Peer Instruction: A User’s Manual in 1997 on this subject [MAZ 97]. Eric Mazur is often considered as the founding father of the flipped classroom, but we cannot forget the American mathematician Salman Khan who was the first to develop the concept and then publish videos on YouTube in 2004 to help his family’s children with math.

34 Marcel Lebrun [BLI 14] thinks it seems more natural to practice the flipped classroom model primary than in secondary school, because primary school education is less compartmentalized than in high school; higher education, for its part, had been less than concerned. However, the emergence of MOOCs revives the debate on the methods of application and added value. 35 Available at: http://Internetactu.blog.lemonde.fr/2013/10/25/lecole-inversee-ou-commentla-technologie-produit-sa-disparition/.

38

Modeling of Next Generation Digital Learning Environments

Since then, classes of this type have flourished on the net. There are now more than “2, 200 mini-lessons”36 on the Khan Academy website. 1.1.12. Synthesis of these evolutions The numerous interactions between pedagogy and technology that have just been described in regards to their “historical” evolution undoubtedly made it possible to enrich teaching–learning processes without forcing teachers to adopt a particular pedagogy. On the contrary, we realize that technologies have not fundamentally changed the way we learn (in the sense of cognitive processes and not teaching methods) but that they have been able to put themselves at the service of (constantly changing) pedagogies chosen by the teachers themselves. Today, we see original pedagogical approaches and new models that feed on technology without becoming a slave to it. As we stated above, following the words of Depover et al. [DEP 09], “ICTs are not [strictly speaking] cognitive tools, but they have a cognitive potential whose use is a factor in the learning and the development of the learner’s skills”. “They allow us, in a formative situation, to build knowledge and activate cognitive and metacognitive learning strategies” [JON 02]. “Improving the quality of learning is the only real objective of what we are presented with today as a technological revolution in the field of education. In order to achieve this goal, educational approaches must be revitalized, based on learning models that make the most of the means at our disposal through technology. In our view, the use of technology is justified only in terms of the relevance of the pedagogical approaches that it allows us to implement” [DEP 01]. 1.2. VLEs: a system of instrumented activity 1.2.1. Virtual learning environments (VLEs) At all levels of the education system, whether in primary, secondary or higher education, it is common to use the term “environment” to define all the elements (human and non-human) that surround the learner and contribute to their cognitive development. If we use the term virtual learning environment (VLE), it is precisely because the main artifacts involved in this development are digital in nature. Oubahssi and Grandbastien [OUB 07] stated that “the concept of the environment in which the activity takes place brings together a set of various resources that can be 36 “Khan Academy”, Wikipedia (2016, https://en.wikipedia.org/wiki/Khan_Academy.

updated

19th

January),

available

at:

A System of Instrumented Activities

39

as much a source of physical or digital knowledge as the tools necessary for the accomplishment of the activity”. However, these environments are also the subject of many competing names when they are “deployed by the training institution: digital or virtual campuses, virtual work platforms, techno-pedagogical environments, etc.; in other words, institutional learning environments” [PER 14, p. 5]. Moreover, the term d'environnement numérique d'apprentissage is mainly used in Quebec, while the term d’espace numérique d'apprentissage (virtual learning space) is more common in Europe. We have nevertheless chosen to adopt the term virtual learning environment because we found that the concept of “environment” was better suited to the systemic nature of our object of study than “space”, which was too often associated with another very different concept which is that of collaborative workspaces (CWS). In addition, an “environment” is defined as “all the elements that surround an individual and some of which contribute directly to meeting his or her needs…”37, or as “the set of conditions […] likely to affect living organisms and human activities”38. These definitions corresponded better to the system we wanted to represent (or model), which justifies our preference for the word “environment” rather than “space”. In addition, the learning environment: “as an empirical object, is most often of a composite nature, articulating numerical and non-numerical elements. As a research object, a learning environment must first be considered as a device within the meaning of our general definition39. However, it is a device responding to a particular configuration with regard to the agents who build them, the purposes they assign to it, the uses it allows, the particular devices that constitute it, etc.” [PER 14, p. 5]. Following Basque and Doré [BAS 98] and in keeping with the constructivist method of learning which is discussed by these authors, Peraya and Bonfils [PER 14, p. 5] defined the learning environment more specifically “as the interaction space in which the learner actively constructs their knowledge by their own experience”, which reinforces us in this choice. However, a worthy qualifier needed to be added to this concept of a learning environment. The question was should we choose “virtual”, “computerized” or “digital”? 37 Larousse dictionary (consulted 5th January 2010). 38 Le Grand Robert de la langue française, Robert, Paris, 2001. 39 From the theoretical point of view, our definition of the device is inspired by Foucault’s work and discursive interactionism [BRO 96]. It considers a device as an “instance of social interaction characterized by its own technological, social and relational, symbolic, semiotic and cognitive dimensions” [PER 14, p. 4].

40

Modeling of Next Generation Digital Learning Environments

Basque and Doré consider that: “The concept of a computerized learning environment encompasses both the idea of the presence of computer resources to support the learners’ approach, the idea of a cognitivist and constructivist vision of learning and the idea of a real or virtual place that houses interacting ‘systems’. And when the place is virtual, it is commonly referred to as a virtual learning environment or a digital learning environment” [BAS 98, p. 40]. We therefore retained this last formulation considering that with the exception of a few rare cases, the “virtualization” of these environments proceeds from digital means. In the English-speaking world, a virtual learning environment is often referred to as a learning management system (LMS). According to the free encyclopedia Wikipedia, “a learning management system (LMS) or a learning support system (LSS) is software that accompanies and manages a learning process or a learning path […]. This kind of computer system offers a collaborative workspace (CWS) including assessment tests that are either subject to validation by the teacher or proposed as self-assessment regulation activities”40. 1.2.2. The principles of the systemic paradigm Larousse41 defines a system (from the Greek σύστημα, or sustema, which means “whole composed of several parts or members”)42 according to the nature of the object that it characterizes (philosophical system, solar system, thermal system, nervous system, etc.). If the VLE is considered an “education system”, the dictionary defines it as a “set of processes, organized practices, intended to perform a function”. In addition to the fact that this definition encourages us to class a VLE as a system, it also has the advantage of firmly expressing its functionality, which our modeling approach will be based on. A system can also be considered “as a set of elements in interaction [dynamic, organized according to a goal]43” [DER 75, p. 92]. In addition to the idea of 40 “Learning management system”, Wikipedia (2017, updated 13th January), available at: https://en.wikipedia.org/wiki/Learning_management_system. 41 Larousse dictionary (online, consulted 8th March 2016), available at: http://www.larousse.fr/dictionnaires/francais/syst%C3%A8me/76262. 42 “system”, Oxford English Dictionary, 3rd ed., Oxford University Press, Oxford, available at: http://www.oed.com/view/Entry/196665, 2015. 43 The phrase between brackets seems to have been completed by Tardieu, Rochfeld and Colletti [TAR 86, p. 32].

A System of Instrumented Activities

41

structure, this definition introduced two complementary notions: “dynamic interaction”, which refers to the idea of the system in operation, and an organization “according to a goal”, which introduces the idea of finality. System theory also distinguishes between closed systems and open systems: “The closed system is defined as being totally isolated from external influences and therefore only subject to internal modifications. The closed system can be defined as an automaton that can only be found in a number of defined states; the knowledge of these states is sufficient enough to define the system”. “The open system is in a permanent relationship with its environment. It undergoes external disturbances which seem to be unpredictable and unable to be analyzed. These disturbances, which occur in the environment, cause system adaptations which bring it back to a stationary state” [TAR 86, p. 32]. In addition, three concepts or formal properties [VON 12, p. 64] apply to open systems. – Totality which considers that the laws governing the behavior of the parts can be established only by considering the place of the parts in total, for example: “If you take any combination of biological phenomena, embryonic development as well as metabolism, growth, nervous system activity, biocenosis, etc., you will always find that the behavior of an element differs inside the system from what it is in isolation” [TAR 86, p. 66]. In other words, “a system does not behave as a mere aggregate of independent elements; it constitutes a coherent and indivisible whole” [TAR 86, p. 33]. – Feedback implies that each element of the system should articulate and interact with the state of the others. “The basic operation of systems is based on the combined play of interactions between system components. To define this set of interactions, simple causal models prove ineffective. In any system where transformations take place, one can identify ‘inputs’ that reflect resources taken from the environment and ‘outputs’ that represent resources returned to the environment. Feedback involves returning the output information to the input system. Feedback loop identification can be used to understand complex system operations.

42

Modeling of Next Generation Digital Learning Environments

The implementation of these loops will lead to a ‘circularity’ of the system. In fact, like a linear causal chain, we can talk about a behavior and an end for systems with feedback, where these terms do not have any meaning” [TAR 86, p. 33]. – Finality which postulates that “in case of disturbance, the system implements forces which oppose it, and which bring it back to the state of equilibrium: these are consequences of the principle of least action” [VON 12, p. 73]. Within the concept of finality, there exists the question of why the system comes into action, what is its purpose? It precedes by far the question of how the system works. Thus, the system allows the employee of a company to make a living, the shareholder to make profits, the researcher to produce knowledge, etc. Some of these purposes are sometimes less obvious than others, such as those related to a need for personal recognition or the conquest of power. This means that the same consequences can have very different origins and that, unlike the closed system, which is completely determined by its initial conditions, an open system can behave independently of these conditions. Applied to social systems, these aims “will be articulated with each other and preserve a certain consistency over time despite environmental pressures. There is self-organization and adaptation of means to ensure the survival and development of the system”44. Jean-Louis Le Moigne [LEM 77, LEM 94] proposed defining a system as “something (anything identifiable), that does something (an activity, function), and that has a structure, evolving through time, in something (an environment) for something (equifinality)”.

Figure 1.5. Drawing of a system [LEM 77, LEM 94]

44 “Approche systémique”, Wikipedia (2016, updated https://fr.wikipedia.org/wiki/Approche_syst%C3%A9mique.

4th

March),

available

at:

A System of Instrumented Activities

43

The diagram in Figure 1.5 clearly shows the reciprocal influences of the evolution, activity, environment and finality of the organization on its structure. In order to put more emphasis on reciprocal influences, a system can also be seen as: “a set of objects and the relationships between these objects and between their attributes. […] Objects are the components of the system, attributes are the properties of objects, and relationships can be understood as what keeps the system together” [TAR 86, p. 32]. In other words, a system can be defined as a complex of interacting elements: “By interaction, we mean elements p linked by relations R, so that the behavior of an element p in R differs from its behavior in another relation R'. If it behaves in the same way in R and R', it does not interact, and the elements behave independently with respect to the relations R and R'” [VON 12, p. 53]. A system can be defined mathematically and in many ways. In his work on General System Theory, Ludwig Von Bertalanffy [VON 12, p. 54] chose to use a system of simultaneous differential equations. Let Qi (where i is a natural integer varying from 1 to n) be any measure of elements pi; these equations will be, for a finite number of elements and in a simple case, of the form: =

(

=

(

=

,

,…,

)

, … ( ,

,…,

) [1.1]

,…,

)

The variation of any measure Qi is therefore a function of all the others, Qi (i = 1, 2, …, n). Conversely, the variation of any Qi results in a variation of all the other measures and the system as a whole. To illustrate the open-system concepts of totality, feedback and finality, Watzlawick, Helmick Beavin and Jackson [WAT 72b] created an analogy with “the family”.

44

Modeling of Next Generation Digital Learning Environments

In a family, they say, the behavior of each of its members is related to the behavior of all others and depends on it. The behavior of the family cannot be reduced to the sum of the behaviors of each of its members. There are system-specific features. This is how they illustrate the principle of totality. When a family is confronted with external events, whether they are significant or not, this can result in a tightening of family ties or, on the contrary, a family breakdown (inheritance, epidemic, etc.). Feedback loops translate into the fact that family responses to environmental disturbances will in turn change the “family system”. Finally, the principle of finality is expressed by the very different motivations that lead each member of the family to behave in this or that way in order to strengthen or find a balance. Behavior in the event may not be very dependent on the initial conditions of the disturbance. “The term homeostasis refers to the existence of feedback mechanisms that serve to mitigate the impact of a change and bring the system back into equilibrium” [TAR 86, p. 34]. It is a principle of moderation, like Lentz’s law in electromagnetism. 1.2.3. The steps involved in the systemic approach 1.2.3.1. The systemic approach in three steps by De Rosnay De Rosnay [DER 75, p. 122], in Le Macroscope, defined three stages of the systemic approach: system analysis, modeling and simulation. They are as follows: – system analysis consists of defining the limits to be modeled, of identifying the important elements and the types of interactions between these elements, and then of determining the links that integrate them into an organized whole; – modeling involves building a model from analysis and synthesis data. The author proposes to establish a complete diagram of the causal relations between the different elements of the different subsystems; – simulation studies the behavior of the modeled system over time, that is, simulating its evolution when one or more variables are modified at a time. To avoid any misunderstanding, it should be emphasized that when we speak of “analysis and synthesis”, it refers to De Rosnay’s definition [DER 75] described above. We must not confuse it with an analysis of the operating system that we will define later. This is a preliminary conceptual step where we seek to represent ourselves, to anticipate the events and to propose actions favorable to the good functioning of the mechanism.

A System of Instrumented Activities

45

1.2.3.2. The systemic approach of Donnadieu and Karsky After De Rosnay [DER 75], Donnadieu and Karsky [DON 02] and then Donnadieu et al. [DON 03] presented the systemic approach not only as a method, but also as knowledge and practice, “a way to enter the complexity”, they say. The process is always carried out in three stages that are almost identical to those described by De Rosnay [DER 75]: “observation of the system by various observers and in various aspects; analysis of interactions and regulatory chains; modeling taking into account lessons learned from the evolution of the system; simulation and confrontation with reality (experimentation) to obtain a consensus” [DON 03, p. 7]. However, it is the tools that these authors integrate into each of the stages that considerably enrich De Rosnay’s approach. Angeline Aubert, Désiré Nkizamacumu and Dorothée Kozlowskidans summarize45 the three steps proposed by Donnadieu et al. [DON 03, p. 7] and describe these tools. Among them, we can mention: – systemic triangulation; – systemic breakdown; – graphical language; – modeling; and – analogy. In the following, we repeat the three stages of the approach proposed by Donnadieu et al. [DON 03, p. 7] by integrating these tools. The first step is exploration. It uses the systemic triangulation method, which involves considering the system according to three different points of view allowing a more in-depth representation of the system “Remarkably adapted to the investigatory stage of a complex system, triangulation will observe this system in three different but complementary aspects, each one linked to a particular observer’s point of view” [DON 03, p. 7]. The aspects of this systemic tool are: – the functional aspect (what does the system do in its environment?), which “is especially sensitive to the system’s purpose or goals” [DON 03, p. 7]; 45 Angeline Aubert, Désiré Nkizamacumu and Dorothée Kozlowskidans, “Agir en situation complexe – Note de synthèse 4 : L’approche systémique”, available at: http://www.esen.education.fr/fileadmin/user_upload/Modules/Ressources/Themes/manageme nt/note_4_approche_systemique.pdf.

46

Modeling of Next Generation Digital Learning Environments

– the structural aspect (its components and their arrangement): “Here we find the analytical approach with a weighty nuance, the emphasis is much more on the relationships between components than on the components themselves, on the structure rather than on the element.” [DON 03, p. 7]; – the historical aspect (evolutionary nature of the system), endowed with a memory and a project, capable of self-organization. Alone, the system’s history will quite often reflect some of the aspects of how it operates. For social systems, it is “through their way of functioning that it is best to begin the observation” [DON 03, p. 7].

Figure 1.6. Systemic triangulation [DON 03, p. 8]

During exploration, it is necessary to identify the different flows, both human and informational, that pass through the system. Applied at this first stage, systemic triangulation takes place by combining these three access paths. More precisely: “More exactly, we move from one aspect to another during a spiral process that allows each passage to gain both in depth and in comprehension, but without ever letting one believe that one has reached the end for this comprehension” [DON 03, p. 8]. The second step is qualitative modeling. It is based on collected information, to develop a faithful and usable map of the system, by visualizing the different interactions between the main components of the system and the environment, as

A System of Instrumented Activities

47

well as the different flows and the piloting actions for system regulation. Systemic breakdown is the second systemic tool, which helps in this case to identify subsystems (modules, organs, subsets, etc.) that play a role in the functioning of the system. Unlike analytical decomposition, we do not try to go down to the elementary components, but “try to clearly define the boundaries of these subsystems (or modules) so that subsequently the relationships that the subsystems maintain with one another will be visible as well as each one’s purpose with respect to the system as a whole” [DON 03, p. 8]. Different standardized schemes have been developed to represent different circuits: organization charts, flowcharts, etc. This is where graphical language comes in as a third systemic tool. “[The authors attribute] four advantages to graphical language: 1) it allows a global and rapid apprehension of the represented system (after learning), 2) it contains a high density of information in a limited space (economy of means), 3) it is mono-semic (a single element or sign) and semi-formal (low variability of interpretation), 4) it has a good heuristic ability (especially for group work).” [DON 03, p. 9] The third step is dynamic modeling: “If the modeler wants their model to be operational, that is to say, allows the user to navigate the complexity and act effectively on it, they must take into account certain criteria and respect certain building laws when constructing their model. Such a process is represented in the diagram below which highlights the four iterative steps essential to any form of modeling. The approach is highly recommended for the study of hyper-complex systems, in particular social ones” [DON 03, p. 9]. Whether the modeling is qualitative or dynamic, it represents here steps to follow. However, modeling can also be considered as a technical process and therefore as a fourth systemic tool.

48

Modeling of Next Generation Digital Learning Environments

Figure 1.7. Systemic approach steps [DON 03, p. 10]

During this modeling stage, a mode of reasoning which continues to permeate researchers’ heuristic approach is analogy. It is seen by the authors as the fifth systemic tool. “In terms of analogy, three levels can be distinguished: – The metaphor establishes a correspondence, often external, between two series of different phenomena or two systems of different natures. Because it is based on appearance, the metaphor is dangerous. Well

A System of Instrumented Activities

49

used, it is precious because it stimulates the imagination and facilitates the creation of new models. – Homomorphism establishes a correspondence between some features of the system being studied with the features of a theoretical model or a simpler real system which is easier to study (which is then called a reduced model). By observations performed on this second (reduced) system, it is possible to foresee certain aspects of the first system’s behavior. – Isomorphism is the only acceptable analogy in a traditional analytical approach. It aims to establish a correspondence between all the features of the object studied and those of the model, where nothing is forgotten or left out” [DON 03, p. 9]. The terms homomorphism and isomorphism are wisely borrowed from mathematical language. Homomorphism (or isomorphism) is an application of one group (or vector space) in another such that it retains its algebraic properties. However, since all isomorphisms are homomorphisms, it is the isomorphism that retains the most properties. If there is an isomorphism between two vector spaces, these are called isomorphic ‒ all the algebraic properties of a vector space are true in the other. It is the same for homomorphisms, but only for the properties of groups, and not for any properties specific to vector spaces. After having defined these three levels, the authors recognize the imperfection of the homomorphic model but see in this imperfection the necessary conditions for all access to knowledge, declaring moreover that isomorphism can only be used for systems of low complexity because it is not useful for complex systems. In the homomorphic model, “the model is without a doubt simpler than the real system, but that’s why we understand it and we can use it to guide our actions” [DON 03, p. 9]. To summarize the steps of the systemic approach described by Donnadieu and Karsky [DON 02] and illustrated in Figure 1.7, these authors specify that in social sciences, the approach does not always need to use all the steps. However, even when limited to exploration, the systemic method thus described remains a good tool for understanding. 1.2.3.3. Elements of synthesis We note that Donnadieu et al. [DON 03] reiterate approximately the three stages of Joël de Rosnay’s approach while complementing it with new tools such as analogy, systemic triangulation, systemic breakdown and graphical language. These are in addition to the basic tools of Joël de Rosnay’s systemic approach, which are the model and the simulation [TAR 86, p. 34].

50

Modeling of Next Generation Digital Learning Environments

From Ludwig von Bertalanffy to Jean-Louis Le Moigne, via Kenneth E. Boulding46, many other variations and sensitivities were noted in the way of designing the various stages of this approach. However, as De Rosnay points out: “[The essential thing is to] locate the systemic approach with respect to other approaches with which it is often confused: – The systemic approach goes beyond and encompasses the cybernetic approach [WIE 48] whose main purpose is the study of regulations in living organisms and machines. – It is distinguished from General System Theory [VON 56] whose ultimate goal is to describe and to encompass, in a mathematical formalism, all systems encountered in nature. – It also moves away from analysis and synthesis. This method is only one tool of the systematic approach. In isolation, it leads to the reduction of a system into its components and into its elementary interactions. – Finally, the systemic approach has nothing to do with a systematic approach, which involves approaching a problem sequentially (one thing after another), being detailed, leaving nothing to chance and forgetting no element” [DER 75, p. 92]. Let us remember that the concept of a complex system is very close to the concept of an open system even if the difference between the two may seem rather vague. Generally, from the moment the system is considered open and consists of a large variety of components organized in a hierarchy, and where these different levels and components are interconnected by a wide variety of links, the system is called complex [TAR 86, p. 33]. Homeostasis also appears as a property or even an essential condition of a complex system. It must be understood as resistance to change. This is one of the most remarkable and characteristic properties of open systems of high complexity. “[A homeostatic system] responds to any change from the environment, or to any random disturbance, by a series of modifications of equal magnitude and direction opposite to those that gave rise to it: these modifications are intended to maintain internal balances” [DER 75, p. 129]. 46 K.E. Boulding [BOU 04a] presents in The Skeleton of Science a nine-level analysis which we will develop and use later. It has largely contributed to informing the conditions of use of the systemic methodology.

A System of Instrumented Activities

51

We will not elaborate here on the concept of a complex system since it will be the subject of discussion in section 4.1. 1.2.4. VLEs seen as open systems The systemic approach is thus opposed to the analytic approach in that it considers the elements of an organization not in isolation, but globally, that is, as integral parts of a set of the different components that are in a dependent relationship. The analytic approach seeks to reduce a system to its elementary elements in order to study in detail and understand the types of interaction that exist between them. This is the case in homogeneous systems, those composed of similar elements and having weak interactions among them. “This applies to phenomena […] which occur in highly mechanized partial systems [VON 12, p. 65]”. It is also observed in biology in a few cases: “The beating of a heart, the nerve–muscle impulse, the action potentials in a nerve are practically the same, whether everything in the organism is studied alone or in isolation” [VON 12, p. 65]. In this case, the principle of summativity47 can be applied. However, this is not at all the case for virtual learning environments (VLEs) whose function, mainly cognitive, is located in an artefactual and social context. Like social systems (of which they are a part), VLEs are made up of elements in dynamic interaction, strongly dependent on each other and the environment. Since they are not totally isolated from external influences, VLEs can therefore be considered as open systems. Several theories and works on human learning by interposed technologies testify to this. However, before mentioning some of these works, let us provide a semantic clarification related to the use of the concept of environment. We use it both to describe and characterize what is outside the “VLE system” and the VLE itself (the “E” of the acronym). This polysemy is likely to create confusion. To be precise, we consider the VLE as the virtual environment (the learning management system, LMS). Considered as a place of learning, the VLE is made up of didactic, pedagogical and technical artifacts. It also virtually integrates the actors involved in training (teachers, tutors, designers, etc.) and the learning subjects as soon as they are enrolled in a training program. Any actor involved in training is physically outside the VLE system and integrated virtually into the system. In the latter case, it is even a subsystem among other subsystems. The environment represents also what is outside the VLE (socio-cultural environment, for example). The traces of the activity are an integral part of the VLE system. The actors of the training or the learners enter it during the preparation or the course of the learning sessions, in the 47 Summativity in the mathematical sense means that the variation of the total system obeys an equation of the same type as the equations of the parts [VON 12, p. 66].

52

Modeling of Next Generation Digital Learning Environments

form of an avatar in a way. They come out once the course is over or the training is over. As we wrote in the journal Cahiers pédagogiques (Pedagogy Papers), the space is virtual, but the learning is very real [TRE 06, p. 51]. Concretely, to avoid any confusion, we will use the acronym “VLE” to refer to the virtual environment (the platform and its contents) and the word environment to characterize the environment outside the VLE system. With these few clarifications, let us now return to the arguments that support the idea that a VLE is a system that exchanges information, matter and energy with the outside world, in other words, it is an open-system environment. According to Baron [BAR 11, p. 111], interest in a systemic approach to education dates back to the end of the 1960s48. Believing that educators have not understood that new means of communication (referring to ICTs) require a radical examination of the teaching–learning process, he advocates: “moving from a teacher-centered learning system to an ‘environment-based’ system, where the relationship between the student and the source of instruction changes, where situations in which learners are active are favored, compared to traditional ones, where they are considered rather passive” [BAR 11, p. 111]. In 1981, the United Nations Educational, Scientific and Cultural Organization published A Systems Approach to Teaching and Learning Procedures: A Guide for Educators. This UNESCO manual presents the reader with: “[this approach] makes it possible to analyse specific educational or instructional situations with a view to making the teaching or learning process more effective. It aims to assist all those professionally involved in education – teachers, headmasters, librarians, adult educators, community leaders, development workers, specialists in teaching methods and techniques, educational administrators, school inspectors, etc. – to identify the basic strengths and weaknesses of existing systems at their working level and to introduce whatever improvements they can decide upon by themselves”49.

48 Baron [BAR 11, p. 11] cites in this connection an OECD report published in 1971 which takes stock of the issue of educational technology, conceived as the implementation of new learning systems. 49 A Systems Approach to Teaching and Learning Procedures: A Guide for

Educators, published in 1981 by UNESCO, Paris, http://unesdoc.unesco.org/images/0004/000460/046025eb.pdf.

available

at

A System of Instrumented Activities

53

This trend in systemic approaches to education has led some researchers to develop systemic approaches to educational planning and development. “Influenced by work on the systemic approach (Lapointe, 1993), he considered a course as a complex system involving a set of elements that should be well planned during a process with the upmost rigor and a search for coherence between the different components of the course (objectives, pedagogical strategies, learning evaluation, media, etc.)” [BAS 04, p. 7]. More recently, Peraya and Bonfils [PER 14, pp. 4–5] have defined a learning environment as “an interaction space in which the learner actively constructs their knowledge, through their own experience” before adding, citing Basque and Doré [BAS 98]: “it is therefore the place that houses a system with its subsystems; this place can be real or virtual”. Donnadieu et al. go even further by claiming: “It is precisely because the system is open that it can maintain its organization, even complicate it, because according to the principle of entropy, a system that does not exchange anything with its environment (a closed or better yet isolated system)50, can only destroy itself (entropic death)” [DON 03, p. 6]. The entropy of a system, symbolized by the letter S, is derived from the Greek ἐν, meaning “inside, in, by” and τροπή, meaning “turning”51 is a concept in thermodynamics. It is defined in this particular area as the measurement of the degree of dispersion of energy (in all its forms: thermal, chemical, electrical) within the system. The second law of thermodynamics (Screated > 0) stipulates that in an isolated system, energy tends to disperse as much as possible. As long as this dispersion takes place, the system “works” (the hot goes to the cold, the nitrogen mixes with the oxygen, the individuals of different origins and cultures exchange their knowledge and work together, etc.) until it reaches a stable state. At this stage, the irreversibility of the phenomenon no longer systematically52 allows a return to 50 Strictly speaking, a closed system does not exchange matter but can exchange energy; an open system can exchange matter and energy; an adiabatic system can exchange work but not heat and finally an isolated system does not exchange matter or energy. 51 Oxford English Dictionary, 2nd ed., Oxford University Press, Oxford, available at: http://www.oed.com/view/Entry/63009, 1989. 52 In our adaptation of the physical concept of entropy in the systemic framework to the social sciences (of education and CIS), we note that the exchange of information can hardly ever be considered as a reversible phenomenon (without creation of entropy). Analogies with irreversible phenomena (melting an ice cube, a mixture of gases) are therefore more relevant than analogies with reversible phenomena (roller coaster without friction). Let us take

54

Modeling of Next Generation Digital Learning Environments

the initial state and certainly not without input of energy: the system does not return to different temperatures, to the separation of two mixed gases or to an original culture. “Such a passage to a higher order supposes a supply of energy; but the energy is supplied to the system continuously only if it is an open system, receiving energy from its surroundings [VON 12, p. 67]”. De Rosnay [DER 75, p. 151] defined this entropic variation as “the irreversible increase of this nondisposable energy in the universe”. Entropy is a very abstract quantity that measures the disorder of the system in question, which can be a country, a company, an organization where someone works, and also a human organism or each of the cells that constitute it. An open system exchanges energy, material and information necessary to maintain its organization to counter the degradation produced by its own operation. At the same time, the system rejects “its entropic residue” in its environment by increasing the entropy of the latter. By opening its borders, it recreates diversity within it, which allows it to increase its complexity and maintain its activity. In fact, by decreasing locally and maintaining its relatively low entropic level, the system is able to maintain its organization in activity and fight against its degradation. According to De Rosnay’s metaphor, “an open system, then, is a sort of reservoir that fills and empties at the same speed, water is maintained at the same level as long as the volume of water entering and leaving remain the same” [DER 75, p. 102].

Figure 1.8. Metaphor of an open system [DER 75, p. 102] advantage of these formal precisions to add, with the same scientific rigor, as for an infinitesimal transformation in a closed system, dS = δSexchanged + δScreated (with δSexchanged = δQ/T and δScreated > 0), whereas in an isolated system, dS = δScreated, because there is no heat exchange with the environment, δQ = 0.

A System of Instrumented Activities

55

Dealing with the “instrumental approach”, Contamines, George and Hotte [CON 03, p. 6], for their part, proposed a systemic analysis centered on the learner. Here, the learner is seen as a subsystem of an instrumented activity system and is prey to the influences of its environment. They also tell us that the anthropocentric approach of Rabardel’s techniques [RAB 95] aims precisely to overcome this dualism of matter/mind to address the cognitive development of the subject by inscribing it in its social and material environment. Moreover, citing Vygotsky’s interactionist theory, they show that learning depends on factors outside the system, by referring in particular to the subject’s cultural heritage, which is external to the subject. In addition, they remind us that “despite the controversial elements surrounding Piaget’s work, the latter agrees, like Vygotsky, that the mechanism of thought and its development is at the crossroads of an outer/inner loop, level/subject” [CON 03, p. 6]. In fact: “[Vygotsky and Piaget] both began to go beyond Cartesian dualism – ‘mind/body’ – to anchor cognitive development and to inscribe the subject ‘in an environment’. For Jean Piaget, the latter is material. For Vygotsky, but also Rabardel and Brassac, it is material and social. The instrumental approach is therefore an approach for which cognition is situated materially and socially” [CON 03, p. 6]. The arguments thus presented each reflect, in their own way, the influence of the physical, sociotechnical and cultural environment on the behavior and learning processes of those involved in VLEs considered as systems. This “permeability” of the boundaries between the system and its environment applies as much to the system as to its subsystems (or parts), which constitute it: hardware, software, platform, tutoring system, organizations, subject’s learners, training actors, etc. Their boundaries (or interfaces) are the heart of interactions between subsystems and between the system and its environment. This set of arguments leads us to consider VLEs as open systems and to treat them as such from now on. 1.2.5. VLEs seen as systems of instrumented activity After being somewhat neglected by researchers for about 10 years, activity theories resurfaced in the 1980s under the impulse of some research on human–computer interaction (HCI). Thereafter, from the 1990s, the new scientific community of CEHL (Computer Environments for Human Learning), which succeeded HCIs, found in these representative theories of any instrumental human action (cognitive and social, instrumented mediation, etc.) with strong theoretical references to study VLEs. In connection with these works, and also with the work of

56

Modeling of Next Generation Digital Learning Environments

the contributors to activity theory, we will show in what way “activity”, and in this case “instrumented activity”, has gradually become the “unit of analysis” for cognition within a VLE. To achieve this, we will seek in particular to overcome the usual categorizations body and mind, affect and intelligence, action and knowledge, intention and motive, means and goals [LIN 02, p. 146] but also subject and object (according to the instrumental approach of Rabardel [RAB 95]) to justify our systemic conception of the VLE. We will do so precisely by considering these different elements in and through their conjunctions. This will lead us to treat a VLE as an instrumented activity system. 1.2.5.1. Vygotsky’s historico-cultural and instrumental approach Despite being a contemporary of Piaget, it was Vygotsky who founded activity theory (AT). He was the first to develop an interactionist theory that makes learning dependent on factors external to the person, that is, social interactions. It is these external factors which, according to the author, model the cognitive schemata that the individual uses to decode and interpret the information that reaches him. Activity is by nature an agent of development insofar as it places the subject in an interrelation with socially elaborate objects and its fellows. In fact, he set “activity” as the unit of analysis to understand and explain human cognitive functioning. For him, “human activity is inserted in a social matrix, in other words in a system of social relations where language plays a preponderant role” [VYG 85]. The activity provides a mediating function between the subject and the world, the others and oneself. Mediation is instrumented, and the instruments are psychological. “They are social in nature and not organic or individual; they are intended for controlling processes of our own behavior or the behavior of others, just as the technique is intended for the control of the processes of nature”. From then on, human thought is constructed during activities with its fellows in a determined social environment. It moves, Vygotsky says, from the social to the individual. This social origin of thought is at the center of his theory. For him, each higher function appears twice during development: first in a collective activity supported by the teacher and the social group, and then in an individual activity. It is therefore in two stages that it becomes an internalized property of the subject’s thought. The learner begins by reasoning with others, then thinking for himself and by himself. This move from the interpsychological to intrapsychological is one of the main ideas of socio-constructivism. “Development can no longer be considered as independent of learning, and learning cannot be considered merely as a ‘private’ relationship between a subject and an object. In this type of approach, we consider that social variables are comparable to the learning processes themselves, and that all development results from learning,

A System of Instrumented Activities

57

thanks to the effect of the inter-individual mechanisms on the intraindividual mechanisms” [ROU 03]. Therefore, contrary to the Piagetian theory that places the development of the individual before learning, that is, as a condition for potential learning (with respect for the stages of development), the Vygotskian theory sees learning as an anticipation of development, provided that the learner is in the zone of proximal development. “If the child takes one step in learning, it takes two steps in its development” [VYG 97]. In addition, Vygotsky was one of the first to describe the relationship between teaching and intellectual development as mediated by “instrumental learning” [VYG 97]. “The instrumental method studies the process of natural development and education as an inseparable whole, in order to discover how, at a certain level of development, the natural functions of a particular child have been restructured” [VYG 85, p. 45]. It is by using “tools” and “systems of signs”, in the first stage where language appears that the intellectual functions develop. Language is primarily a means of communication in the social relations of the individual. It is only in the second stage, by transforming itself into an inner language, that it becomes an instrument of thought. “Learning occurs in a cooperative social situation, when the semiotic devices used by the interaction companion can be incorporated by the learner” [CRA 01, p. 149]. It is therefore when engaging in an activity that the individual learns to master the tools he encounters, to turn them into tools of thought by transforming them into action schemes. His whole theory could be summarized as follows: “Teaching is a process of cultural transmission that leads to the development of mental abilities, not yet mastered by subjects, and that they build by learning specific tools that make up human works (literary, scientific, artistic, etc.)” [AMI 03]. It is for this reason that we also mention a historico-cultural and instrumental approach when we talk about the Vygotskian theory. 1.2.5.2. Léontiev’s developments In the wake of Vygotsky, Léontiev sought to study the psyche of the individual by studying their activity, which in his opinion results from the subject’s superior (and no longer only biological) need to perform actions. According to him, the result

58

Modeling of Next Generation Digital Learning Environments

of the activity would be the result of actions undertaken by an individual going in the direction determined by their purpose. Leontiev [LÉO 81], quoted by Jermann [JER 96], distinguishes in this theory three interactive levels of relations between subjects and objects: activities, actions and operations. The activities are intentional, “in close relation with a conscious goal, a motivation” [JER 96]. Each motive is related to a need for an object (material or ideal) to be satisfied by the subject. They “may give rise to a multiplicity of actions” [JER 96] that are generated to satisfy that purpose (the object). For example, an activity may consist of repairing an automobile, the object being the repair of this vehicle. The individual (the subject) performs a certain number of actions (in our example, buys the mechanical parts, transports them and assembles them) in order to reach this goal. The subject assisted by a set of tools (or artifacts) that serve as a mediation between the subject and the object, tools that include not only material objects but also symbols, signs and languages used by the subject (e.g. negotiating the price of the vehicle to transport the spare parts). Action can be defined as a process functionally subordinate to activity, directed by its purpose. “Actions have both an intentional, oriental aspect (what is to be done) and an instrumental aspect (how to do it: an anticipated plan and a general method to reach intermediary goals)” (Linard, quoted by [JER 96]). They are performed by operations that are compiled and unconscious procedures. An action may serve several activities. The action of driving a vehicle, for example, can be used to transport passengers, to move materials or to prove driving abilities during a car race. While the action puts the activity into effect while responding to a genuine need and a real object, the operation divides the action into sequences to facilitate learning. To drive an automobile, operations are necessary: shift gears, accelerate, brake, etc. The operation often finds no justification in itself and its only “object”, its only motivation, is the appropriation of the action concerned (driving). Operations can take various forms in various situations: partial executions of actions, etc. An activity is associated with a project, an action with a strategy and an operation with conditions necessary for its execution. “Actions and operations are in a dynamic relationship that allows an action to become an operation […]. As actions become operations, the subject can take action on higher level actions. When the execution conditions of an operation have changed, it can again obtain the action status to be specialized and adapted to these new conditions” [JER 96].

A System of Instrumented Activities

Dimensions

Functions

Activity

Project (motive, reason, intention)

Incentive (incentive function)

Action

Strategy, design, purpose

Orientation

Operation

Conditions of realization (means and processes)

Realization

59

Table 1.1. Léontiev’s stages of activity [LÉO 81]

1.2.5.3. Overview of activity theories and the systemic approach We have just recalled in the preceding parts the main principles of the systemic approach as well as the foundations of the activity theory in order to set a theoretical reference framework for the study of any instrumented human activity which takes place inside a VLE. We therefore open this new section with the intention of showing that a VLE cannot only be considered as a place that houses instrumented activities, but also, in a more encompassing and formal way, as an instrumented activity system. For Peraya and Bonfils: “Any device53 [whose authors agree to recognize their systemic nature] instrumentalizes human activity [RAB 95], and its analysis must therefore be considered as ‘inseparable’ from the analysis of activity. Learning, communicating, working, and producing alone or together must be considered as instrumented activities […]. Consequently, this definition constitutes a framework of analysis which aims to account for a wide variety of concrete, empirical devices (a training system, a classroom, a digital campus, videoconferencing software, etc.) as well as their degree of granularity and complexity (e.g. a digital workspace, a personal learning environment or a simple ‘tool’ for chatting)” [PER 14, p. 4–5]. In reference to the degree of granularity and complexity evoked by these authors, Basque and Doré [BAS 98, p. 40] considered “a teacher and learners as a system: each individual being a subsystem or component whose actions are oriented towards the development of new knowledge”. For these authors, “the environment is likely

53 From the theoretical point of view, our definition of the device is inspired by Foucault’s work and discursive interactionism [BRO 96]. It considers a device as an “instance of social interaction characterized by its own technological, social and relational, symbolic, semiotic and cognitive dimensions” [PER 14, p. 4–5].

60

Modeling of Next Generation Digital Learning Environments

to host many of these activities, in addition to the tools and equipment needed to achieve them” [BAS 98, p. 40]. We first of all retain that the systemic approach opens up an original and promising way to observe and study instrumented action at the heart of a VLE. “Constantly combining knowledge and action, the systemic is presented as the indestructible alliance of knowledge and practice” [DON 03, pp. 1–2]. In the words of the researchers cited, we also note the desire to definitively integrate the concept of system of actions and activities by incorporating the tools and equipment needed to achieve them. Actions and activities are also seen as oriented towards a common goal, in this case learning (a teleological property specific to both systems and theories of activity). Finally, there is a marked interest in the diversity of the constituent elements of the system (the variety of subsystems). It comes back as a leitmotif under the argument that it is precisely this diversity that allows for, ultimately, maintenance of the system’s activity. What we see as fundamentally common between the systemic approach and the activity theory is precisely that these theories consider, on the one hand, that the system and the activity are directed towards a goal, and on the other hand that these two theories apprehend the studied phenomena in a conjunctive and not disjunctive way. We have just made some arguments about the teleological (goal-oriented) aspect of the activity theory and of the systemic approach. In particular, the conjunctive or all-encompassing nature of the systemic approach is self-evident as it is one of the characteristics inherent to systems. The holistic (conjunctive and integrative) nature of the theories of activity to confirm this kind of “homomorphism” between the systemic approach and activity theory remain to be seen. Monique Linard was one of the first to perceive the systemic nature of these theories: “[These theories], emerging from the human sciences, are beginning to be recognized for their power of explanation and understanding of human–technical relationships. They reject the dualistic split between body and mind, affect and intelligence, action and knowledge, intention and motive, means and goals. They form the interaction between subjects, objects and environment, the dynamic basis of construction of intelligence, psyche, language and meaning. They are based on the principle, ignored by the rationalist paradigm, of the evolutionary genesis of the structures and mental functions of individuals within their social relations, from their interactions with their physical, psychological and socio-cultural environment” [LIN 02, p. 146].

A System of Instrumented Activities

61

By rejecting this dualist split, Monique Linard resolutely enters into a conjunctive logic that links elements and variables instead of separating or isolating them. Moreover, Rabardel’s instrumental approach, based on the idea that an instrument is the result of the use of a tool (artifact), highlights the interactions between the subject and the tool during an activity. For him, “an instrument is not only a material entity but a bifacial entity, composed of an artefactual component and a schematic component [in the sense of Piagetian schema]” [CON 03, p. 6]. Instrumental genesis refers to the process of development and evolution of the instrument during the activity. The tool (artifact) provokes in the user the desire and the need to learn how to use it (instrumentation). In return, the use made by the subject changes the tool (the artifact) which is assigned new functions other than those originally planned by the designer: this is what Rabardel calls the process of instrumentalization. These permanent interactions link the tool to its user. It is only by considering the subject and its tools in an indissociable way, that is, conjunctive, that it becomes interesting to give an account of their evolution. “The instrumental approach, like generally constructivist approaches to cognition and instrumented activities, tends to abandon the usual categorization of subject/object. In this regard, Jean Piaget argued that the confrontation ‘… of science ultimately leads us to highlight what the analysis of each particular knowledge stresses from the outset but to varying degrees: the close interdependence of the subject and the object …’ (Piaget, 1970). The activity becomes the unit of cognitive analysis, just as it is the case for Vygotsky (Vygotsky, 1997) and his contemporaries who developed the activity theory” [CON 03, p. 9]. By thus characterizing the activity theories, these researchers undeniably confer on the instrumented activity a systemic character that we will translate, by a system of instrumented activity. 1.3. Conclusion The systemic approach of a VLE naturally leads us to extend this representation to its activity by referring to the theories of the same name. We have seen that this activity was resolutely goal-oriented (system-specific) within an instrumented organization. By instrumented organization we mean a cognitive organization dealing with technical objects, such as software, for example, which mediate and script the activity. These objects “are means of action for men, i.e. instruments of their actions” [RAB 95]. According to Rabardel and Bourmaud [RAB 03, p. 665], they are also “components of more general systems that integrate and go beyond

62

Modeling of Next Generation Digital Learning Environments

instrument systems”. Each instrument of this system is linked to its user by the concomitant processes of instrumentation and instrumentalization (of instrumental genesis). This particular reliance54 is but one example of the connective nature of the relations between humans, activities and techniques. Action and knowledge, intention and motive, means and goals are others. And this list is far from exhaustive. Thus, the teleological and conjunctive characters specific to activity theories can only encourage us to recognize in them their resolutely systemic nature. To conclude this chapter even more comprehensively, let us say that the systemic approach “is not only knowledge, but also a practice and a way of entering into complexity” [DON 03, p. 7]. To better understand how a VLE works and thus to guide the action of any decision makers (political leaders, training managers, experts, researchers, etc.), we had to clearly identify the object to which we were referring to in our research. That is to say: can a VLE be seen as a system of instrumented activity?

54 The concept was originally proposed by Roger Clausse (in 1963) to indicate a “psychosocial (information) need: reliance on isolation”. It was retaken and re-elaborated in the late 1970s by sociologist Marcel Bolle de Bal who linked it to the media. “To the notion of connections, the reliance will add the meaning, the purpose, the insertion into a system” [BAR 04, p. 5].

2 Modeling Instrumented Activity at the Heart of the Virtual Environment

2.1. Introduction When we are confronted with the complexity of a phenomenon, we are usually led to use a model that allows us to record our perceptions using specific tools to facilitate analysis and understanding. Once this is recorded and analyzed, the phenomenon can be brought back to its original context. As Paul Valéry said, “we access our reasoning only through the models we have built for ourselves” and “we only could reason on models” [VAL 75, p. 835]. Modeling is therefore the representation of one system by another, facilitating understanding. It can be considered as “a technical process which makes it possible to represent an object or a situation or an event deemed complex, for the purpose of knowledge and action”. It is used in all scientific fields concerned with complexity. However, modeling is also an art in which the modeler can express his vision of reality when it may not necessarily be the reality. In this respect, the systemic modeling of complexity refers us straight back to constructivist epistemologies. “For constructivism, knowledge is constructed by the modeler who has the project in his permanent interactions with the phenomena he perceives and conceives. This process of building active knowledge is at the heart of the process of modeling phenomena or complex perceived systems” [LEM 99, p. 23]. “The same reality, perceived by two different modelers, will not necessarily lead to the same model” [DON 03, p. 9]. According to this constructivist conception of learning, knowledge of the phenomenon studied is constructed through the elaboration of the model: the Modeling of Next Generation Digital Learning Environments: Complex Systems Theory, First Edition. Marc Trestini. © ISTE Ltd 2018. Published by ISTE Ltd and John Wiley & Sons, Inc.

64

Modeling of Next Generation Digital Learning Environments

modeler is the one who learns from his model by seeking meanings for the phenomena he models. In this sense, the modeling approach is opposed to the empiricist and positivist conceptions of knowledge production, which claim that this can only be produced by repeated observations (inductive approaches). Karl Popper’s falsifiability theory [POP 94] shows the limits of inductive methods (as well as deductive ones for that matter). “The latter has devoted his work to undermining a deterministic conception of the universe and society by drawing a conclusion that directly contradicts the Hegelian conceptions that remain predominant in the West: ‘the future is open’. There is no predetermined meaning and the future will be what we decide to do” [ROC 98]. “It is this production of intelligibility in a perceived complex situation that justifies […] the ambitious project of Systems Science: to develop conceptual languages that aim at understanding meaning rather than explaining form” [LEM 94, p. 278]. Paul Valéry wrote: “We have always sought explanations when it was only representations that we could seek to invent” [VAL 75, p. 837]. Having outlined the epistemological foundations of systemic modeling, it is now necessary to present the various models likely to meet the stated need. To do this, it is possible to group them according to the function they perform (design, analyze, evaluate, etc.), or according to the levels of investigation chosen by the modeler (micro, meso and macro) or according to the outline of the system to be modeled (training device, teaching scenario, resources available, etc.). It is also possible to classify them in chronological order as all these models have evolved together with the educational systems they are supposed to represent. In France, for example, Baron [BAR 94] “explains how they have moved from classical computer-assisted teaching systems (CAT) to intelligent computer-aided instruction (ICAI), and then to virtual learning environments (VLEs)”. He adds that “an additional development occurred in our country at the end of the decade with the introduction of the concept of an intelligent tutoring system (EIAH1). These were still in use in 2011, with greater interest in collaborative learning and distributed training systems” [BAR 11, paragraph 26]. These forms of mediation have also been described in different theoretical frameworks (activity theories, complex system theories, etc.). Other possible categories can therefore be envisaged according to this new dimension.

1 The acronym EIAH comes from the French concept of Environnements informatiques pour l’apprentissage humain that we have translated into English as “Intelligent Tutoring Systems”.

Modeling Instrumented Activity at the Heart of the Virtual Environment

65

Having mentioned these different regrouping possibilities, we have chosen to present the few models which we will have to discuss by classifying them first according to their function and then according to the outline of the system to be modeled. We will end this chapter with an overview of the most representative models in their category and a description of the modeling languages they use. 2.2. Modeling instrumented activity, yes, but why? 2.2.1. What type of model are we talking about? Legendre [LEG 93] distinguished between two main modeling categories: the object model and the theoretical model, which respond to different needs. The object model is seen, according Mouloud, Jaulin, Tonnelat, Goguel, Guinand, Boudon, Richard, Victorri and Damasch (1999), cited by Harvey and Loiselle [HAR 09, p. 96], as a “concrete model, constructed from experimental data, which reflects as accurately as possible some of the properties, geometric or functional, of the object and laws to which it is subject”2. We refer to this type of model in particular by expressing our intention to model the instrumented activity at the heart of a VLE. Thus, “it becomes possible to organize knowledge into more easily communicable models, then to use some of these models in reflection and action […] to enable the organization of knowledge and make action more effective” [DER 75, p. 92]. For its part, the theoretical model “allows us to elaborate, from the object model, a theory that transforms the studied phenomenon to a more general phenomenon (concept) in line with experience and confronted with it” (op. cit. by Harvey and Loiselle [HAR 09, p. 96]). In this case, to model the instrumented activity of a VLE would involve identifying invariants, general principles, both structural and functional, that can represent them as a whole, and the goal would be to move from individual cases to generalization. For Backer [BAC 00], the models respond to two essential functions that Desmoulins and Grandbastien remind us of [DES 06a, p. 161]: – “to propose a conceptual framework allowing the designer to represent the world, in particular with a view to speculating on its behaviour”; – “to provide representations that can be manipulated by humans or by machines that are useful for their activity”.

2 This definition of a “concrete model” is borrowed from l’Encyclopaedia Universalis, “Modèle: Les grands articles”, Chapter 3, “Le modèle en biologie”, 2016, France.

66

Modeling of Next Generation Digital Learning Environments

Thus, modeling the instrumented activity at the heart of a VLE amounts to describing the functional needs and solutions in a system of representations (often visually and graphically). These functional needs and solutions make up the design and implementation phases of a VLE project. They can also be used to evaluate or analyze the activity which exists or for understanding, control or even prediction purposes. Many models exist to achieve one of these goals. Some of them have been designed exclusively to fulfill one of these functions, while others allow us to fulfill several of them at the same time, such as those to guide engineering and to evaluate, analyze or control the system’s behavior. Let us start by describing these models capable of performing several usage functions before switching to models exclusively created for the design or analysis of a VLE.

Figure 2.1. Predictive scenario and descriptive scenario [PER 04a, p. 412]

2.2.2. Models for multiple uses 2.2.2.1. “Design” models for device analysis Some VLE design models can also help evaluate or even analyze its activity in the context in which it occurs. The modeling of a learning scenario, for example, in an IMS-LD type EML language, can help the modeler to describe the expected or actual course of a learning situation with the aim of ensuring the appropriation of a precise set of knowledge [PER 04a, p. 410]. These two authors describe the following two scenarios and their objectives: ““A predictive scenario” is a scenario established a priori by a designer in order to set up a learning situation, whether or not it is instrumented by digital technologies. The definition of a predictive scenario can pursue several complementary objectives:

Modeling Instrumented Activity at the Heart of the Virtual Environment

67

– “streamlining design”, assisting designers in defining learning situations and providing them with methodological guides. Designers can be both specialists involved in an industrialization training process and teachers or trainers who have to modify their practices in a more individual way; – “improving the efficiency of learning situations”, in particular by enabling the actors responsible for setting them up and monitoring them to have an explicit framework to better orient learners towards the activities to be carried out; – “empowering learners”, by making the learning objectives and the structuring of activities they have to accomplish clearer. This approach is particularly used in self-training contexts or in projectbased pedagogical approaches for which log sheets are provided; – “rationalizing learner assessment”, by having a means of measuring the differences between the learner’s actual activity (or group of learners) and that described in a typical scenario defined a priori. This type of approach is based on the behavioral theories of learning. “A descriptive scenario” is a scenario describing a posteriori the actual progress of a learning situation, including in particular the traces of the activity of the actors and their productions. The use of descriptive scenarios can serve different purposes: – “conducting a didactic evaluation of a learning situation”, using learning events and traces encountered in real-life situations to infer or verify assumptions about the actual appropriation of knowledge. This approach is frequently used in the field of experimental psychology; – “assisting in the assessment of learners”, by analyzing all traces collected and possibly comparing them with a predefined ideal model, contributing to the constitution of profiles, allowing us to individualize learning. […] The IMS-LD specification is clearly in the context of a predictive scenario where it is a question of describing a learning situation a priori. It should be noted that the proposed implementations insist on arguments for streamlining design and evaluation, most often in a framework of industrial training” [PER 04a, p. 412]. 2.2.2.2. “Activity analysis” models for device design Conversely, it is possible to improve design approaches based on business analysis models. Engeström’s model, for example, may in some cases facilitate this design while it is initially designed to analyze it. Bjørklia, Røedb, Bjellandb, Gouldc

68

Modeling of Next Generation Digital Learning Environments

and Hoffd [BJØ 07, p. 171] showed how this model has been useful for designing a ship’s navigation system (specifically, automated steering control). Starting from the description of the navigation, they could create this specific tool. In their study, Engeström’s model “provided them with a framework to describe the people involved in the ship’s navigation, their use of tools, and what governed their behaviour in a coherent way” [BJØ 07, p. 176]. In addition, it provided them with a basis for a design language that was used to design a tool (course and track pilot) which was intended to replace the helmsman. “This study argues that the use of activity theory can make an important contribution to a design process” [BJØ 07, p. 176]. Moreover, Jonassen’s work [JON 00] bears witness to this when he revisits activity theory as a framework for the design of student-centered learning environments. 2.2.2.3. Summary Thus, many of these models cover the whole life cycle of VLEs (analysis, design, use, feedback, reconceptualization, etc.) and ensure, depending on how we use it, that it carries out one or more of these functions. 2.2.3. Models with specific uses 2.2.3.1. Modeling to design a virtual learning environment There are many models that meet this need and which we could refer to as instructional design (ID) or learning design [BAR 11, paragraph 26]. This choice reflects the importance of the prior modeling stage in the development cycle of this type of environment3. In his book The Conditions of Learning [GAG 68] (first edition published in 1965), Gagné paved the way for instructional design where the purpose is to study, design, build and adapt teaching devices, training or courses. Instructional design is therefore what is commonly called pedagogical engineering in France. Pedagogical engineering mainly deals with the development of pedagogical scenarios, or training engineering, which are supposed to promote learning. “In educational sciences, the development process of a product is designated under various names: pedagogical design models, learning

3 To be more precise, David Merrill “distinguishes the ‘development’ of instructional systems from instructional design theory. The latter must, according to him, have three components: two ‘descriptive’ theories (knowledge and teaching strategy) and a ‘prescriptive’ theory connecting knowledge to be acquired and strategies to use in order to foster learning, composed of rules of the type: if (learning of such knowledge), then (such strategy)” [BAR 11, paragraph 19] (http://mdavidmerrill.com/Papers/papers.htm).

Modeling Instrumented Activity at the Heart of the Virtual Environment

69

systems, engineering systems, development model, development method or development sequence” [HAR 09, p. 97]. These authors note that the multiplicity of designations surrounding research and development sometimes causes some confusion. Citing Richey and Nelson [RIC 96] they highlight in particular “the ambiguity that exists between the pedagogical design model and those of research and development. They point out that the presence of such a fragmented terminology gives way to a multitude of interpretations causing confusion that often reduces understanding” [HAR 09, p. 97]. We note that the instructional design (ID) is “an area of English research interested in the prescriptive aspects of education design, that is, how we can organize it so it is as effective as possible” [DES 06b, p. 137]. The latter takes a detailed look at the different terminologies in summary note form. It also explains the function and place of the “model” in the field of instructional design: “An instructional design model would be a practical application of the theory, and thus would be used rather to construct examples of certain conceptions of teaching or learning (e.g. behaviourists, constructivists). This is why ID models are rarely tested: they are not supposed to reproduce, ‘describe’ part of the reality, but ‘prescribe’ a procedure and be effective. We have preferred, for the sake of simplicity, to use the term ‘model’ uniformly here rather than method [he says]. Putting aside the notion of a formal model, describing a process in a mathematical language, a consensus seems to have been reached about the idea that a model is a simplification of a phenomenon that we want to study, a simplification that favors the observation, description, or prediction of the phenomenon. [This definition is consistent with that of object model (i.e. reduced model), inspired by the ‘hard sciences’]. So, we are talking here about participating objects, according to Simon (2004), a ‘science of design’, or an engineering, i.e. a discipline showing how things ‘should be’. And the models [that the author describes in this summary note] are simplifications of the concept of teaching phenomenon, rationalizations in order to make it more effective” [DES 06b, p. 139]. The models described by Philippe Dessus in his summary note make it possible to develop teaching, that is, to prescribe the manner in which it must be organized so that it is as effective as possible. It is also important to emphasize that these models are neither theoretical models in the sense that Legendre [LEG 93] understands

70

Modeling of Next Generation Digital Learning Environments

them, nor object models, such as those we are interested in: they are above all methods or descriptions of procedures. A whole field of research opened in the 1980s, bringing together mainly engineers and some researchers in the humanities, to think about how to design architectures that are both ergonomic, efficient and easy to use for the purpose of human computer interactions (HCI), whether it is interactions between humans and computers (interactivity) or between humans (interactions) using computers. In doing so, researchers in the humanities are increasingly investing in this field and are joining computer scientists and researchers in computer science to create a transdisciplinary field of science, which aims to study how to develop computer artifacts to accompany or even to improve learning and teaching processes in computer environments. At the beginning of the 2000s, this trend was reflected in the progressive approximation of the issues and approaches of the scientific communities participating in the “Hypermedia and Learning” conferences and the “Interactive Computer Learning Environments” days. As part of ATIEF4, these communities decided to make this approximation a reality by merging these two conferences into one, which is called “Intelligent Tutoring Systems” (EIAH) [DES 03]. The first of these took place in Strasbourg in April 2003 and again in June 2017. This transdisciplinary scientific field is both an empirical object (a learning environment) and also a research object that now bears the same name ITS: “[And] the term ‘conception’ of an ITS refers to imagining, thinking, elaborating and representing a computer artefact, taking into account the pedagogical objectives pursued and the constraints of various natures that may be exercised in particular targeted pedagogical situations” [TCH 09, p. 1]. The purpose of ITS research work then consists of “studying the computerized pedagogical situations and the software that allows these situations to occur” [TCH 09, p. 1]. Then, with the development of programming languages, we have gradually seen the development of graphical representation languages whose function is specifically to facilitate the design of this type of environment. Procedural (not to say theoretical) models of instructional design such as “linear, behavioural, knowledge-centered, cognitive, learning-centered, and constructivist models, taking into account the world of work” ([DES 06b, p. 138] cited by Baron [BAR 11, paragraph 38]) are increasingly using graphs to represent the procedures and sequencing of actions. Designers thus need visual models to fully understand the system they want to build. As Seel (1997) quoted by Dessus [DES 06b, p. 138] states:

4 French Information Technology Association for Education and Training.

Modeling Instrumented Activity at the Heart of the Virtual Environment

71

“The theory/model relationship in the field of ID is different from that commonly accepted in the field of philosophy of science. In the latter, a model (theoretical, according to Seel) is a local version of a theory, whereas in ID, only the theory requires empirical proof: An ID model would be a practical application of the theory and would be more likely to be used for the construction of instantiations of certain conceptions of teaching or learning (e.g. behaviourists, constructivists)” [DES 06b, p. 138]. These instantiations can also be represented graphically to facilitate the designer’s approach: these are then diagrams or visual models. “Diagrams serve as support for problem solving and performing complex tasks. They allow us to locate data, to determine the goal to reach as well as the constraints to be respected, then to construct a solution and to execute it” [PAQ 02a, p. 34]. Graphic and visual modeling is at the heart of the educational engineering method. According to Paquette: “It achieves four goals: – it visualizes a system as it is or what we would like it to be; – it specifies the structure and behavior of the system; – it obtains plans and specifications that will guide the construction of the system; – it subsequently documents the decisions we have made” [PAQ 02a, p. 42]. These graphic models are numerous and have evolved over time, with cognitive psychologies and with the resulting conceptions of teaching. This is concluded by the idea shared by Baron [BAR 11, paragraph 38] that there has not been an evolutionary rupture “but rather adaptations to the models so that they remain compatible with what one thinks about when we refer to the real activity of teaching”. We will give some examples of VLE design models and discuss a possible paradigm shift in the art of designing and modeling the latest generation virtual learning environments such as MOOCs. To end this section, we refer readers interested in to a detailed description of the various research and development models from the field of education to the article by Harvey and Loiselle [HAR 09], which is accessible online.

72

Modeling of Next Generation Digital Learning Environments

2.2.3.2. Modeling to analyze a virtual learning environment Previously seen VLE design models can also be used to analyze instrumented activity once the device or interface is designed and implemented. However, there are also specific models that are exclusively dedicated to the analysis of training devices, used in particular to judge the adequacy that can exist between the objectives set by the designers or teachers and their more or less effective implementation. From a teaching scenario created by a teacher or a designer for example, it becomes possible, thanks to these models, to undertake a comparison of its real activity with a predetermined activity in a model, then to compare the student activity to equally predetermined activities, and finally to compare these activities to the pedagogical approach chosen by the teacher. The model then helps the teacher or designer to confirm that these three aspects are consistent. Models have been developed and tested for this purpose by several researchers, such as those proposed by a collective at the Université de Sherbrooke led by Desrosiers [DES 13]. These models are based on “works by well-known authors in post-secondary education, such as Christian Barrette’s meta-research (2011, 2009, 2005), Rolland Viau’s (2009) model of academic motivation, and the ‘added value’ of Docqs, Lebrun and Smidts’ team (2010)” [DES 13, p. 3]. These models allow, in particular, the analysis of pedagogical practices mediated by socio-constructivist and learning-centered technologies while promoting academic motivation. It would be much too time consuming and certainly useless at this stage to review all the analysis models of instrumented activity that have been proposed in recent decades as there are too many. However, if there is a model that has really marked this intention at the beginning of the century in the EIAH research community, it is undoubtedly Engeström’s [ENG 87], which has become a reference model. Its use in system analysis will be widely discussed later and will be the subject of a thorough “contradictory” study. This model, rooted in activity theory, places it in a train of thought that best reflected learning phenomena by interposed technical systems in the 2000s [BEL 96]. In this model, a family of more or less explicit rules mediates the relationship between the subject and the community; the “division of labor” mediates the relationship that is established between the community and the object of the activity. From a methodological point of view, it is the behavior of each of the constituents of the activity system and the tensions or contradictions that appear within it that have made it possible to understand the activity’s evolution. These contradictions are observed at different levels: primary (within the various poles themselves), secondary (between the different poles), tertiary (between similar activities involving different motives), and finally quaternary (falling within networks of similar, i.e. central and related, activities).

Modeling Instrumented Activity at the Heart of the Virtual Environment

73

This model of analysis allows us to consider activity as a basic unit of human development. The activity aims to transform the environment while transforming those who are involved with it. This transformation is done using instruments that connect people to the world of things and to other people. By his activity, man is enriched by all the experience of humanity. 2.2.3.3. Modeling to evaluate a virtual learning environment The evaluation will be characterized by the use of research methods specific to the education sciences in order to appreciate the effects of these learning environments considered as teaching or training devices. It consists of viewing the application of a particular teaching method as “the testing of an implicit or explicit hypothesis which underlies it” [NIO 82, p. 36]. If, for example, a VLE teacher proposes collaborative work to prevent students from dropping out, it is because he makes the assumption that this way of working determines what causes dropouts and non-dropouts. The evaluation will therefore consist of ensuring that this is the case and that the dropout level is the source of these results. There can be numerous evaluation objectives. Most often, it is to assess the success or failure of the implemented strategy. The complexity of the situations faced by VLEs can then lead us to use explanatory methods specific to evaluative research. However, evaluation should not be limited to the experimental model. It is also necessary to try to place evaluation in the broader perspective of the analysis of the systems of activities. Choosing action processes as units of analysis would be a good start. In terms of activity systems, this would consist of moving from so-called “summative” evaluations to “formative” evaluations [SCR 67, NIO 77]. The “summative” evaluation in pedagogy is the one that, at the end of the teaching, records the pupil’s acquisitions. It reports the student’s level at the time of the assessment to the institution and the parents. Therefore, this is a statement that offers nothing in return and so there is nothing formative in this evaluation. The “formative” evaluation is confused with the teaching act. It adds a progressive element to the educational action (in the sense of expected progress). This is done through the permanent back and forth between the evaluation and remediation processes that are implemented immediately afterwards. “It allows the student and the teacher to periodically assess the level and pace of the acquisitions as well as the reasons for this level and pace. This then allows teaching methods to be adapted. This evaluation makes it possible to go beyond the ‘objectives-results’ model” [NIO 82, p. 55]. Beyond the assessment, it will be possible to assess the success or failure of the VLE as an instrumented activity system. This evaluation will allow learning “if it exhibits and explains the unintended effects of programs, if it acts as revealing

74

Modeling of Next Generation Digital Learning Environments

of preferences by bringing out confused or hidden goals, if it brings to light the implicit theories behind practices” [NIO 82, p. 55], in this case pedagogical ones. Another way to evaluate a virtual learning environment is to compare it to a model [DAM 08]. This author distinguishes three main dimensions for evaluating an interactive system: utility, usability and aesthetics. For him, the utility, which he subdivided into several criteria, is the most important of these dimensions. Indeed, an interactive learning system with low utility would not meet or only partially meet the training objectives. Thus, utility could be subdivided into different observables: functional capacity, system performance, reliability, content, quality of support, scenario and metaphors.

Figure 2.2. Evaluation dimensions, adapted by Dambreville [DAM 08] according to [SEN 93]

2.2.3.4. Modeling to anticipate the emergence of new behaviors To be able to develop plausible scenarios of how a system may behave in the future (in this case a VLE), we need a model that, based on the data it can provide, makes it possible to analyze certain factors of uncertainty. These will be cross-referenced (or compared) with heavy tendencies, attractors, leading to the development of plausible scenarios. Le Moigne [LEM 99] reminds us: the incompleteness of a model (uncertainty factor) “will not constitute a regrettable

Modeling Instrumented Activity at the Heart of the Virtual Environment

75

imperfection, but rather a condition necessary for the anticipation, by simulation, of a possible emergence of behaviors within this complex system”. It is obviously not a form of fortune telling which, with all related omens, discovers what is hidden in the future. This relies on the data provided by the model to identify emerging phenomena (or emerging properties). For example, a plausible scenario may result from a consistent combination of answers to questions raised by factors of uncertainty. Then, if one is confronted with contrasted scenarios, a more detailed analysis can be carried out in order to highlight the similarities and the discrepancies of each one of them. However, we must remain cautious about the conceptual framework in which these scenarios are usually built. There is, indeed, a debate among specialists in the field on the status of prospective practice that we cannot pretend to ignore. “If we abstract from the infinite variations attributable to this or that individual, we can say that, since the beginning of the 1970s, two tendencies have been opposed: the supporters of a ‘prospective-attitude’ and those of a ‘prospective-science’. For the former, in the continuity of the founders of the French School of Foresight (Gaston Berger and Hugues de Jouvenel), prospective is a particular ‘look’ focused on the long term and which draws upon the need for methods and techniques and acquired social sciences. For the latter, prospective is itself a science, a branch of the social sciences” [TOL 97]. That said, we would rather be on the side of the supporters of a “prospective-attitude”, but this question does not seem to require an immediate stance from us in this debate. What is certain is that our approach is not militant; it does not seek to make prospective a science. 2.2.3.5. Modeling to personalize the learning environment As Lefevre and Molinari recall in a framework for a call for projects5 for ORPHÉE RDV from January 2017: “The question of personalizing learning environments can be approached from different angles as well as in various educational contexts. It can relate to distance learning, academic or professional training, and can be implemented in environments as varied as

5 Call for submission of the workshop “Personnalisation et adaptation dans les environnements d’apprentissage” within the ORPHÉE RDV framework of January 2017.

76

Modeling of Next Generation Digital Learning Environments

hypertexts (e.g. websites), intelligent tutors, serious games, MOOCs and other online courses. It is aimed at all learners including learners with special needs, learners working alone or learners working in a group. It also meets multiple pedagogical objectives including the promotion of autonomy and self-regulation of learning. Personalization can include the contextual recommendation of resources, the use of appropriate support and the consideration of the learner’s emotional and cognitive state. This customization can be done by several actors (the learners themselves, the teaching teams), and the proposed approaches vary leaving the control of the personalization sometimes to the learner, sometimes to the teaching staff, and sometimes to artificial intelligence techniques”. Moreover, while VLEs offer new ways of learning, new approaches and new learning contexts, they are still required to provide the learner with pedagogical pathways and teaching content adapted to their needs and to their profile. The development of adaptive teaching systems aims to meet this objective (see section 2.4.3). 2.3. Contour, components and hierarchical levels of a system of instrumented activity Modeling a virtual learning environment is first of all a case of drawing the outline of a world that we wish to study or construct, “whether it is a particular reality whose components it is important to recognize, an acting reality whose actions must be identified, or even a future reality whose evolution must be determined” [LEM 08, p. 104]. We have provided a description of the components that constitute this world in the foreword of the previously mentioned book [TRE 08]: “it brings together human beings grappling with technical, pedagogical and didactic artefacts, intended for the cognitive organization formed of a wide variety of interacting components or elements that are further nested and organized into hierarchical levels” [TRE 08, p. 9]. However, typically, “a complex system does not suffer from fragmentation into finer underlying systems. It is indivisible, if fragmentations are necessary, it is to be able to delimit the outlines of its components” [TRE 08, p. 35]. 2.3.1. Perimeters, objects and components Given the diversity of today’s virtual learning environments, modeling can range from relatively “simple” objects (such as instructional software) to more complex

Modeling Instrumented Activity at the Heart of the Virtual Environment

77

objects such as long-distance learning, remote open training, virtual campuses or even MOOCs which include all the pedagogical and non-pedagogical functions of a training program, a faculty or university platform, or even a national training platform (e.g. FUN). In addition, these objects or devices very often introduce multimodal pedagogical approaches such as blended learning, hybrid courses and open and distance learning, etc. They are built “according to an audience’s (and possibly an institution’s) objectives and specific working conditions” [POT 03, paragraph 81] and implement human and material resources. All these multimodal devices require a hierarchical approach to teaching methods or techniques. Therefore, we must take the necessary step back to properly identify the system to be modeled. This step back will allow us to be in the right position to observe and investigate the system. At the higher level is appropriately “multimodality”. The following device6 terms are attributed to it: mixed architecture or sometimes hybrid paths. “Multimodality (if we can call it this), falls into a category of its own in our hierarchy of pedagogical modalities or techniques” [DEN 13, 30th July]. It covers the three lower levels: macro, meso and micro. 2.3.2. Three levels (macro, meso, micro) associated with business processes Unsurprisingly, this hierarchization brings us back to the three levels of the use of the term “device” identified by Demaizière [DEM 08, paragraph 19] and the associated business processes: that of instructional design (learning engineering), the construction of pedagogical sequences (pedagogical engineering) and the design of resources or specific tools. “[Demaizière himself recognizes] that it is difficult to find a unifying concept or perspective behind the term ‘device’ as the uses vary […]. From the tool considered in isolation or presented in a training sequence from the globalizing point of view of the training engineer, we find the term at these three levels, considering each of them in a favored way in one of the perspectives according to its culture” [DEM 08, paragraph 19]. The macro level is therefore what is known as learning engineering. It deals with a learning modality (distance, presence or experiential), says the co-founder and associate director of C-CAMPUS.

6 Remember that, “as a research object, a learning environment must firstly be considered as a device according to our general definition. In any case it is a device that responds to a particular configuration with regards to the agents that construct them, the purposes they assign to it, the uses it allows, the particular devices that compose it, etc.” [PER 14, p. 5].

78

Modeling of Next Generation Digital Learning Environments

“These three modalities correspond in fact to three major pedagogical families that can be characterized according to their pedagogical context. The first two are part of a formal learning environment where the primary intention is to learn. Conversely, the third group includes informal learning methods. Learning is only a side benefit of the activity or the work situation” [DEN 13]. The objects involved in modeling are in this case all those which make up one of these modalities placed in a social, cultural, economic and political environment. “The term macro specifies that this is the top-level category of teaching techniques” [DEN 13]. The meso level corresponds to pedagogical engineering. It concerns the strategies implemented to facilitate learning. “Each learning modality or macro pedagogy is divided into pedagogical activities. These include role plays, presentations or quizzes for face-to-face training, serious games, e-learning, social learning, e-reading for distance learning, mentoring, tutoring, sponsoring or good practice sharing groups for experiential training. These activities may involve digital technologies” [DEN 13]. Finally, the micro level focuses in particular on the nature (text, sound, video), the structure and the quality of courses or the effectiveness of tools (quizzes, etc.). “In recent years, with the introduction of e-learning and the renewal of experiential training, the level of pedagogical activity has now become insufficient to effectively create training programs. Development needs to be much more micro and more detailed activities need to be imagined. In a typical course, this involved presentations, case studies or practical work, and that was sufficient. With blended learning, we must create e-learning modules integrating micro-activities which are a few minutes long. Examples included watching a video, interactive graphics or a mini quiz. With tutoring, we are led to design courses incorporating microtechnology such as duplication, feedback, flash explanation and learning missions, so that all these activities or techniques qualify as micro-pedagogy” [DEN 13].

Modeling Instrumented Activity at the Heart of the Virtual Environment

79

Figure 2.3. Macro, meso and micro techniques [DEN 13]

2.3.3. An example of distance learning device analysis We measured the importance of a good relationship between these three levels (macro, meso or micro) as part of research conducted in 2012 at the University of Strasbourg [TRE 12b, TRE 14]. This was a comparative study of the various distance learning devices used in all parts of this university at the time of the proposed merger of its founding institutions (grouping of the three founding universities of ESPE and of IUT). A research report available on the IFE website (French National Institute for Educational Research) [TRE 12b] reports on this work. It also allowed us to note the importance of outlining the object to be studied: the quality of the analysis seemed to depend on it. In this example, there was nothing to prevent us from studying the distance learning devices of each founding university independently before making the comparison. However, we quickly realized that, in doing so, we would not be able to make a prediction of the overall behavior of the unified device, as the reciprocal influences between the subsystems (founding universities) had consequences for the general system (all universities together). As Lemire wrote: “any fragmentation loses its properties to the system, they all emerge only if the system is grasped in its entirety. These properties, in many cases, exist only because of the relations which exist between the whole and its parts” [LEM 08, p. 34]. A study on the industrialization of distance education, conducted at the same time and in this context [TRE 12b, TRE 14], also showed us how much political

80

Modeling of Next Generation Digital Learning Environments

discourses of legitimizing strategies implemented (ideologization)7 influence the evolution of the overall device [TRE 12b, TRE 14, TRE 16b]. Here again, this influence could not have been observed if we had not studied jointly, during the same period, all the devices implemented. In doing so, we found, for example, that the unifying discourses of “politics” had more impact on the choice of tools to be retained and on the teaching strategies to be favored for the new university than 10 years of practical experience. 2.3.4. Associated levels, objects and models At each level of investigation and each type of object studied corresponds to a model which aims to represent the activity at the heart of the environment. In the following section, we describe some “object” (and not “theoretical”) models as defined in section 2.2.1. Prior to that, we propose here two more refined classifications of “object” models. Generally, when variables are few, we usually use simple analogical models. An analogical model is an object model that reproduces as closely as possible the phenomenon observed by using hardware objects to represent the system. The model of an erupting volcano is an example. Observation of the behavior of the model makes it possible to learn from the phenomenon (the displacement of the lava flow in our example). However, as soon as the number of variables increases, the analogical model quickly shows its limits. This is particularly the case for more complex systems such as the latest generation VLEs. This is where the mathematical models come into play: we refer in this case to numerical modeling. It consists of constructing a set of mathematical functions describing the phenomenon. By modifying the initial variables, it is then possible to predict the modifications of the physical system. On the contrary, Harvey and Loiselle define two other types of models within the field of education sciences: local models and general models. “Local models are characterized as such when they are developed in a specific way for a particular discipline or for specific learning in that discipline [DEP 02]. When applied to teaching situations, these local models make it possible to link development directly to specific content. According to Depover and Marchand [DEP 02], this contributes to a conceptualization better adapted to learning mechanisms, but also to the establishment of more precise 7 In addition to rationalization and technologization, ideologization is the third criterion of the industrial dimension of training [MŒE 98, p. 22].

Modeling Instrumented Activity at the Heart of the Virtual Environment

81

prescriptions as to the way in which teaching should take place in this particular context. These authors also highlight that we speak of global models when models highlight common learning strategies that would be valid for various disciplines. As they have a more generic character, these models can also be described as ‘general’ [SCH 95]” [HAR 09, p. 97]. 2.4. Summary of models and modeling languages In the field of ITS, modeling approaches are numerous and globally refer to the needs stated in section 2.2, namely to design, analyze, evaluate and customize these environments, and finally anticipate their future, even if points of convergence or overlapping areas still exist. Each of these approaches corresponds to models that we have also characterized and categorized in this same section and in section 2.3.4. We will now focus on these models, in particular their languages. More specifically, we will start by mentioning some design or development languages of pedagogical engineering models, in other words models of pedagogical design. Although the polysemous nature of this concept is very often discussed, a converging point of view lies in the fact that it “refers to the set of theories and models for understanding, improving and applying teaching that promotes learning” [PQA 02a, p. 111]. We will gladly refer the reader interested in semantic precisions concerning these concepts to the article by Harvey and Loiselle [HAR 09] which “dissects” all the polysemic aspects. We will then take a step back to look at, from a more macro point of view, learning engineering models, which include the pedagogical and non-pedagogical functions of a VLE. Then, after having quoted some adaptive models, we will look at some activity analysis models, in particular those of Engeström. 2.4.1. Models of pedagogical engineering (EML) According to Paquette, the pedagogical engineering of Education Modeling Languages (EML) “must be seen as a methodology supporting the analysis, design, implementation and planning of learning systems usage, integrating the concepts,

82

Modeling of Next Generation Digital Learning Environments

processes and principles of instructional design, software engineering and cognitive engineering”8. The concept of a learning object is based on the methods and techniques of pedagogical engineering. There are different models manipulating this concept: LOM9, SCORM, EML, etc. For Pernin and Lejeune [PER 04a, p. 409], this learning object concept consists of three classes of objects: learning units, activities and resources. The LOM adopts the IEEE’S definition: “A learning object is any entity, digital or non-digital, that can be used, reused, or referenced during technology supported learning”. This definition, considered too broad from a technical and scientific point of view by Wiley [WIL 00], cited by Pernin and Lejeune [PER 04a, p. 410] and Koper [KOP 03], is quoted in the context of digital technologies by the same authors: “any digital, reproducible and addressable resource, used to perform learning activities or learning support activities, made available for others to use”. However, for Jenni [JEN 09, p. 20], there is still confusion between the concept of learning objects and the concept of resources and in particular that of educational resources. For her, “learning objects” clearly refer to a succession of “techniques” ensuring not only the assimilation of all the data (encapsulation), but also the support of the interactions and the management of access. On the contrary, the educational resources refer to the content and correspond to a resource in its most basic meaning10 [JEN 09, p. 20]. “Teaching resources correspond to everything related to teaching and learning in their entirety. They refer in particular to educational content. The materials are created, selected and edited specifically for learners and teachers. These resources can contain both text and multimedia. They are not necessarily limited to simple reading material” [JEN 09, p. 20]. We will then distinguish the learning objects from these educational resources (content objects).

8 Paquette G. “L’ingénierie cognitive du téléapprentissage”, in Taurisson and Senteni (eds.), L’apprentissage collaboration, Québec University Prs, cited by Pernin and Lejeune [PER 04a, p. 407]. 9 LOM specification, Learning Object Metadata; objectif: indexer des objects pédagogiques pour les reutiliser dans des curricula, available at: https://ieee-sa.imeetcentral.com/ltsc/, and in archived version: http://archive.wikiwix.com/cache/?url=http%3A%2F%2Fltsc.ieee.org% 2Fwg12%2Ffiles%2FLOM_1484_12_1_v1_Final_Draft.pdf 10 This refers to the definition by Petit Larousse 2010; the resource is equivalent to “means available” (p. 884).

Modeling Instrumented Activity at the Heart of the Virtual Environment

83

In addition, the new conceptions of learning that prevail in the minds of designers and which predominated in the early 2000s are mainly centered on the learner’s activity. They have profoundly changed the way VLEs are created, as well as the pedagogical engineering models themselves. It was a case of moving away from old beliefs. These were beliefs that learning would be reduced to a simple “quasi” mechanical transfer of knowledge between a pedagogical resource filed online (content object) and the learner’s cognitive system. However, the aim was not to question the value of having content objects in specific virtual places, but rather to consider that these pedagogical resources did not represent the only key to success in a learning process, with the learner’s activity being at least as important as the teaching resource in this process [KOP 01, p. 3]. We have thus moved from a “content-centered approach” to a “process-centered approach” as defined below by Pernin and Lejeune [PER 04a, p. 410]. Each of these approaches includes specifications11 which are aimed at standardizing exchanges in the field of online training. There are, for example, specifications that model the resources (ARIADNE, Can Core2, Dublin Core, LOM, etc.) and others that model learning activities (EML, IMS-LD, etc.). “[The content-centered approach] is directly related to the increased opportunities offered by the Internet to access vast amounts of information, particularly of an educational nature. It highlights the advantages of the computer object approach in promoting new uses based on the principles of sharing, reuse and aggregation. Standardization work in the field has resulted in the LOM specification, which defines a set of metadata for indexing learning objects for cataloguing and reuse and proposes a successive aggregation object model ranging from basic documents (an image for example) to very high-level entities (a curriculum)” [PER 04a, p. 410]. “[The process-centered approach is interested in] defining pedagogical engineering methods capable of ensuring the establishment of ‘resources and pedagogical means’ facilitating the design and implementation of training” [PER 04a, p. 410]. New modeling languages (EML) are born from this desire to combine resources and pedagogical means. An EML is defined by the CEN/ISS as “a model of information and semantic aggregation describing the contents and processes involved in a learning unit from a pedagogical perspective and with the aim of ensuring reusability and interoperability”.

11 https://www.imsglobal.org/metadata/index.html.

84

Modeling of Next Generation Digital Learning Environments

The learning unit can be considered as a gestalt [KOP 01, p. 3], that is, as “a perceptual entity … treated by the subject as a whole rather than a juxtaposition of parts”12. In practice, it can be of different kinds and sizes, it can be a course, a curriculum, a workshop, a practice and a lesson. To relate this concept to the “multimodality” levels that we defined in section 2.3.2, the learning unit would be at the macro level. Koper himself says that: “A learning unit can be represented by: – online learning (completely across the web); – blended learning (mix of online and face-to-face learning); – hybrid learning (mix of different media: paper, web, e-books, etc.)”13 [KOP 01, p. 4]. At a lower level (meso) is the pedagogical scenario. For Paquette: “[Describing an instructional scenario consists] of describing the activity or learning and support activities, the resources required to carry out the activities and outputs that should result from them […]. As part of a learning unit, the pedagogical scenario consists of a ‘learning scenario’ and an ‘assistance scenario’” [PAQ 02b, p. 444]. Let us quickly go back to these two last concepts. “The ‘learning scenario’ represents the description of the expected or actual progress of a learning situation aiming at the appropriation of a specific set of knowledge by specifying the roles, activities as well as knowledge manipulation resources, tools and services needed to implement activities” [PER 04a, p. 410]. “[The ‘assistance scenario’ includes] the activities of trainers or other types of facilitators, the resources to be used, the productions to be made and the rules and instructions for intervention with the learners” [PAQ 02b, p. 444]. To illustrate this definition, Paquette [PAQ 02b, p. 444] cited and explained some possible assistance scenarios: the scenarios of methodological assistance, assistance by questioning, assistance by presentation, tutorial assistance and assistance by analogy.

12 According to Larousse. 13 Translation by author.

Modeling Instrumented Activity at the Heart of the Virtual Environment

85

It is possible to literally describe a pedagogical scenario with standardized words related to a field of activity or using tables with several columns. However, these are often difficult to use because the courses are complex, often nonlinear or offering very different pedagogical modalities (see section 2.3.2). Therefore, it is often easier to describe them graphically. As we have already stated, this is the case for graphic or visual languages. A modeling language can therefore be represented by graphic symbols describing the model’s components represented and the links that exist between them. This type of language is a good communication medium for the design team and its partners. They also make it possible to give a global vision of the training envisaged and to make all the components of the scenario coherent: objectives, learning and support activities, methods, educational resources, means and services, etc. Let us begin by briefly presenting the EML language and then one of the most popular educational modeling languages that follows: the IMS-LD language. 2.4.1.1. The EML language It should be recalled that initially, the first approaches from different consortia (ARIADNE, IEEE/LTSC9) were only related to the description of resources [OUB 07]. From there, Rob Koper [KOP 00] proposed to describe learning situations using a pedagogical modeling language in order to define and materialize: – the relationship between the intended objectives (in terms of knowledge, abilities or skills); – the actors involved in learning processes; – the activities carried out; – the content needed to put them in place. He also proposed describing actual learning situations using a language that positions learning situations, not resources, at the center of the process. “This world materialized during the specification of a first language, that of EML (Educational Modeling Language) which largely inspired IMS Learning Design specification in 2002” [PER 04a, p. 410]. The notation system used to describe the units of study uses an internationally recognized metalanguage: The Extensible Markup Language (XML)14. XML, in turn, is based on the ISO SGML standard. XML 1.0 became a W3C recommendation, 14 The XML language is a language that can describe data using tags and rules that can be customized (Ludovic Roland, author of the Openclassroom XML course: “Cours XML”, available at: https://openclassrooms.com/courses/structurez-vos-donnees-avec-xml/qu-est-ceque-le-xml).

86

Modeling of Next Generation Digital Learning Environments

the World Wide Web Consortium, on 10th February 1998. Using SGML and XML, it is possible to develop a vocabulary specific to the chosen domain. This was then very useful to describe the units of study. The first basic structure of EML is shown in Figure 2.

Figure 2.4. The structure of an EML [KOP 00, p. 25]

This representation can be represented by an XML tree structure. Figure 2.5 gives an overview and Figure 2.6 a simplified translation (without content) in XML. 2.4.1.2. The IMS-LD language IMS Learning Design is a modeling language that takes into account a wide range of pedagogies for online learning. Based on the same principle as EML developed by Rob Koper [KOP 01] at the Dutch Open University, this generic, flexible and independent language of implementation systems has been designed to allow teachers to choose from among the many pedagogical approaches that exist the one that best fits their goals and the teaching–learning situation they seek to implement. In other words, this modeling language “provides a framework for taking into account the diversity of pedagogical approaches while ensuring the exchange and interoperability of learning materials and learning units staging them” [PER 04a, p. 410]. For the reasons mentioned, the IMS-LD model is very similar to the EML model, but it nevertheless has some differences: instead of the “unit of the study”, it uses the concept of a “learning unit” and also uses the concept “resource” instead of “object”. Finally, an activity cannot only use resources but can also produce new ones [OUB 07].

Modeling Instrumented Activity at the Heart of the Virtual Environment

Figure 2.5. EML shown in an XML tree structure [KOP 00, p. 25]

87

88

Modeling of Next Generation Digital Learning Environments

Figure 2.6. Simplified notation of a unit of study without content

The IMS-LD language has been specified by the IMS-Learning Global Consortium15 after an extensive review and a comparison with other languages in a wide range of pedagogical approaches and their associated learning activities. This review led to the creation of the Learning Design V1.0 specification in February 2003 (IMS-LD). “[This language] provides a conceptual framework for modeling a learning unit and aims at proposing a good compromise between the genericity allowing us to implement various pedagogical approaches and the power of expression allowing a precise description of each learning unit” [BUR 05].

15 https://www.imsglobal.org/learningdesign/index.html.

Modeling Instrumented Activity at the Heart of the Virtual Environment

89

More specifically, the IMS-LD language allows us to share and reuse scenarios regardless of the environment where we want to implement them. It makes it possible to formally specify the progress of a learning unit (module, course, session, etc.). To illustrate its potentialities, let us take at random an example of e-training given by Burgos et al. [BUR 05]: “Teachers want to migrate their lesson plans to an online system. They create specific content in: TXT, PDF, PPT, DOC, AVI, XLS, HTML, RTF, SXW or other file formats but also Internet links. They can incorporate as many resources as desired, create documents and link them together, along with assessments, additional information about educational goals, prerequisites and more. They invest time in modeling everything and preparing their courses so that these formats are used properly. They can decide to create the structure with HTML pages in order to be seen by an Internet browser. This solution is useful if they simply want to show documents and if they are addressing learners browsing freely in the course pages. If they decide to incorporate a control over the methodology, the evaluation, the answers to the questions, the rights of ownership or inscription for example, they must necessarily insert the structured content in a system that allows all of that. To do this, a learning management platform (LMS) and a virtual learning environment (VLE) are the best options. In this way, they maintain the information, the course that has migrated, transcribing paper on the screen with some additional possibilities mentioned above. It is possible to add forums, online chats, communication services, hide or show information, depending on the level of expertise or the user’s profile, etc. Two aspects have been addressed so far: content/resources and activities/functions built on content/resources. The former is an unfinalized product that we can use in different applications and use whenever we need it. The latter come from the pedagogical-didactic approach, completely related to the tool used to model them. If we succeed in keeping this second category as independent as possible from the first one, we will not need to rebuild it in case we change the content, change platforms, or update the application. We can change the content of the resource itself, but its link remains and the structure works: that’s how a specification comes into play” [BUR 05].

90

Modeling of Next Generation Digital Learning Environments

Figure 2.7. Life cycle of creating a learning unit [BUR 05]

In addition, IMS-LD uses a theatrical metaphor to define the structure of a learning unit: “A play is divided into one or more acts and is performed by several actors who may take up different roles at different times in the play” [BUR 05]. For each role, a number of activities must be carried out to complete the learning process. An activity is located in an environment including services (chat, forum, messaging, etc.) as well as resources (objects of knowledge or content). In addition, all roles must be synchronized at the end of each act before dealing with the next one [BUR 05]. IMS-LD also offers a choice of level modeling (Figure 2.8) to define “prescriptive” scenarios (level A), “personalization” scenarios (level B) and “dynamic” scenarios (level C). The so-called prescriptive level A describes the “method” used, with all learners assumed to be following the same course. “Level A includes the definition of ‘method’, plays, acts, scores, learning activities and tutoring environments. The key of the specification is that it contains the description of the elements that

Modeling Instrumented Activity at the Heart of the Virtual Environment

91

configure IMS-LD and the coordination between them. For example, role-parts define the activities that are carried out in a role to complete an act, and subsequently, a play” [BUR 05].

Figure 2.8. Architecture of the IMS-LD specification, level A, B and C16

Level B allows learning pathways that may change during the execution of the learning unit depending on the decisions that can be made during the activity. “Level B adds properties, conditions, tutoring services and elements acting on the set to Level A. It provides specific means for creating complex structures and learning experiences. Properties can be used as local or global variables, storing or removing information for a single user, a group, or even all people involved. Through these mechanisms, the learning path can change during the execution of the unit and decisions can be made by taking into account dynamic aspects. Level C adds notifications to level B, that is, for example, an e-mail sent and a function shown/hidden are linked to a specific activity, depending on how the previous activity was performed” [BUR 05].

16 IMS, Learning Design Specification, Boston, United States, 2003.

92

Modeling of Next Generation Digital Learning Environments

Figure 2.9. Graphical model editor LD (G-MOT)

From an operational point of view, we have just seen (Figure 2.7) that the design of a scenario involves the use of a publisher. On the one hand, this must allow a description of the activities specific to learning and assistance, and on the other hand, the resources required to carry out the activities and productions that result from it. In order for this description to be clear and easily understandable, only graphical representation seems to combine these qualities. This type of representation must also be easily transposed into a model that complies with the IMS-LD standard, a model that makes the represented scenarios reusable and exchangeable. BF Conseil et Formation17, in partnership with LICEF, proposes the

17 Source: BF Conseil et Formation is a specialist in tailor-made services in training engineering and educational engineering. The name of their website, “IngeGraph”, represents the contraction of two words which seem to be antinomic, “ingérie” (engineering in French) symbolizing complexity, and “graphic”, evoking an easy-to-understand visual representation. Available at: http://www.ingegraph.com/.

Modeling Instrumented Activity at the Heart of the Virtual Environment

93

publisher G-MOT that meets these requirements. It can be downloaded from the LICEF website18. Figure 2.9 gives an overview.

Figure 2.10. Example of a “classic” scenario in compliance with level B IMS-LD

In Figure 2.10, we give an example of a graphical representation of a “classic” learning scenario by presentation and control of knowledge using a knowledge test while respecting level B’s IMS-LD standard (choice of path according to the test results). Note that red ellipses represent activities (functions), “raised rectangles” represent resources and “the little men” represent actors. “P”-type links indicate the order of the scenario flow and I/P-type links indicate the use or production of a resource.

18 http://poseidon.licef.ca/gmot/launch.jnlp.

94

Modeling of Next Generation Digital Learning Environments

Figure 2.11. Example of XML codes of the specification [BUR 05]

Once the scenario is complete, simply save it as an “IMSmanifest” file. From a technical point of view, the specification defines an XML document which is no different than an imsmanifest.xml file (see Figure 2.11) describing a very detailed teaching scenario that links the actual resources in each format with it [TAT 03, p. 4]. Once the IMSmanifest file of the pedagogical scenario has been created, it will have to be completed using the “Reload LD Editor” which will make it possible to produce the final (zipped) package whose function is to collate linked resources, weblinks and many learning materials and services. “[The package is therefore] a compressed file with: (a) an XML ‘manifest’ that describes the ‘method’, the play, acts, roles, activities,

Modeling Instrumented Activity at the Heart of the Virtual Environment

95

environment, properties, conditions and/or specification’s notifications, which also indicates the resources related to it; and (b) the group of documents or resources mentioned in the XML ‘manifest’” [BUR 05]. “[An XML file] is completely different from an HTML website which points to the same resources […]. An XML ‘manifest’ is not just linked content as in an HTML structure but it is a single folder that: (a) indicates the contents and resources whereas an HTML page is a resource in itself which can also contain references to other resources; (b) which combines learning structure and process with a unit of learning while an HTML website is a series of linked and/or structured web pages with no underlying learning ‘method’; and (c) which can provide conditions, properties, tutoring services and notifications. Such a file makes it possible to adapt the access to resources according to the users’ actions and the data exchanges coming from the interactions whereas an HTML website is a source of passive and static information” [BUR 05]. In conclusion, the scenario editor allows teachers to model their teaching scenario as they represent it (they can then return to it if necessary). In the back office, this editor generates a scenario model conforming to a formal model (a metamodel). This formalized scenario is then carried out by a scenario player, that is, a motor capable of ensuring a link with the virtual environment, to instantiate19 the scenario and allow its online execution in connection with the virtual environment. For example, the imsmanifest.xml integration in Moodle goes through the “add an IMS Content Package” resource. However, in its specialized packages version, there is a folder including a manifest integrating a Learning Design in place of the traditional “organization” part.

19 “In object-oriented programming, we call an instance of a class an object with a behavior and a state, both defined by the class. It is therefore an object constituting a copy of the class […] Instantiation is the act of instantiating, that is, creating an object from a model”, “Instance (programmation)”, Wikipédia (2014, updated 17th February), available at: https://fr.wikipedia. org/wiki/Instance_(programmation).

96

Modeling of Next Generation Digital Learning Environments

Figure 2.12. IMS-LD learning unit

In conclusion, the steps to follow are as follows: – create a teaching scenario graphically; – transpose it according to the IMS-LD standard; – implement it in Moodle 2.X, thanks to the IMS CP resource. A recent procedure was proposed by Lepage [LEP 15]20 for Moodle 2.8. 2.4.2. Training engineering models In general, training engineering refers to approaches and methods of designing training devices21, which are supposed to respond in a rational and efficient way to the objectives intended by the designers. In this section, we obviously deal with instrumented activity devices, in other words e-training, distance learning devices, 20 http://docplayer.fr/4480373-Moodle-2-8-3-ajout-d-un-scorm-en-mode-manifeste-scormexterne.html. 21 According to the meaning already repeated many times.

Modeling Instrumented Activity at the Heart of the Virtual Environment

97

distance learning or even open and distance learning (ODL). We will not revisit these different acronyms and prefer instead to refer the interested reader to a research report that states these terms [TRE 12b, p. 84]. When necessary, we will use the term ODL to represent this set. The engineering process generally takes place in three successive stages: analysis, design and implementation22. 2.4.2.1. Analysis The notion of “analysis” in the field of training engineering has a particular meaning that differs from that of the “analysis” of the teaching process in the computer environment. To avoid any confusion, let us say that “the analysis of the teaching process” is often presented in a systemic framework, the system considered being constituted of three essential poles: the “knowledge” pole, the “pupil” pole and the “teacher” pole – see, for example, Brousseau23 [ART 90, p. 5]. Changes in this general pattern subsequently appeared to better mark the presence of the computer tool in teaching. Some authors even add other poles to the system to emphasize the fact that this computer tool specifically affects other system components. Engeström’s “systemic” model [ENG 87] offers, for example, and according to this same meaning, its own framework of “analysis”; its poles are indeed different (subject, tool, objects, rules, community, division of labor) and refer to a different theoretical framework (AT). “Analysis” in the field of training engineering corresponds to the (traditional) study of the feasibility of the project, which involves the analysis of the need and the establishment of the service functions (of use and estimation). It aims to identify the training organization or company, the context of the application, its situation, what it expects, the target audience, its objectives, content elements, forms of evaluation (diagnostic, normative, formative, summative), as well as the different constraints (temporal, budgetary, etc.). This analysis generally leads to the drafting of specifications and/or a model of expression of the need and functions that the system must ensure. The “use of case diagrams” specific to object-oriented design models, for example, illustrates what can be established in this case.

22 Aït Hennani, 2008, from University of Lille 2, details each of them in one of his presentations from which we have also taken some elements. Source: http://iut.univlille2.fr/fileadmin/user_upload/documents/Emplois/IngenierieDeLaFormation_Mohamadia_0 6juin2008.pdf. 23 [BRO 86].

98

Modeling of Next Generation Digital Learning Environments

2.4.2.2. Design The actual design will consist of a description of the context, the training objectives, the pedagogical objectives (general and specific), a general description of the device, the detailed training program(s), the pedagogical methods used, the sequencing of each of the training sequences and sessions, the description of the target population, the trainer(s) profiles, the assessment device and schedule of the device. We note that at this stage of the device design, we integrate both pedagogical aspects (like obviously the pedagogical engineering seen beforehand in section 2.4.1), and also non-pedagogical aspects (technological, economic, organizational, decision-making, and sometimes legal). 2.4.2.3. Implementation Training is a project that requires a steering mechanism composed of a steering committee which is in charge of the management of the training project; a project manager who manages the people involved in the training project; a project team that implements the training project; and a network of resource persons who participate in one way or another in the training project. Its implementation also passes by the animation of the implemented device, animation which is essentially educational; communication between the various actors must be facilitated. The proper flow of information is therefore essential at this stage. The evaluation of the device is the final stage of implementation: it gives an idea of how the training works. It must be conducted at different levels and regularly and repeatedly to validate the effects. 2.4.2.4. Conclusion Training engineering includes the methods and practices of demand analysis and training needs; the design of a training project; the definition of methods and means to be implemented; coordination and monitoring of training; the evaluation of the training as well as the modes of validation envisaged. Training engineering should thus be distinguished from pedagogical engineering (which refers to practices) or engineering of professionalization, which is based on the alternation between formal or informal learning situations. As Oubahssi and Granbastien [OUB 07] point out, most engineering proposals focus primarily on the learner’s activity, which represents only a partial view of the overall activity, or focus only on a given category of actors, and not on the whole of a device of instrumented activities. From this observation, several modeling attempts were undertaken based on business models in general and within global open and

Modeling Instrumented Activity at the Heart of the Virtual Environment

99

distance learning (ODL) processes. These models cover more broadly the activity of all the actors involved in the training process and make it possible to particularize it for each phase of the process [GRA 03]. The interest lies not only in this desire to want to model an entire device, but also to take into account all the players acting on the system. They also wanted to “ensure better interoperability of data related to these activities between the different software components used in an e-ODL” [OUB 07]. This activity model, which more broadly covers the activity of all actors in the training process, has been published in Oubahssi et al. [OUB 05]. The educational activities naturally find a predominant place in this model: elaboration of the pedagogical modules, preparation and integration of the contents, tests and simulation, collaboration and co-operation with the other actors, etc. But what distinguishes it in particular from the previous ones lies in taking into account new activities which also make new actors appear. For example: “Each of the managers and administrators uses his environment to carry out the following activities: – administrative management activities: management of user accounts (learners and teachers), management of training group accounts, management of schedules, management of training agreements…; – technical management activities: data security, maintenance of smooth training, document management…; – training management activities: definition of domains, definition of disciplines, definition of training levels, document management…” [OUB 07]. Figure 2.13 shows a class diagram, derived from object-oriented design models, which details the activity model proposed in an open and distance learning environment. “At each stage of the process is associated an environment in which the actors perform one or more activities. In this model, an ODL environment is therefore composed of a set of work units, links, rules, and resources. In the work units, activities take place. The work unit is defined as a composition of activities performed by a set of actors in a given ODL environment […]. Each activity is characterized by a set of prerequisites, objectives, and is defined by a state (for example, in progress). The environment in which the activity takes place makes it possible to group together a set of resources of all types and the tools necessary for accomplishing it.

100

Modeling of Next Generation Digital Learning Environments

Each activity uses and produces a set of resources (tools, services, results …). The main actors who manipulate the activities are: author, counselor, tutor, learner, evaluator, administrator, general administrator and pedagogical administrator. Rules represent a set of conditions or constraints that allow the smooth running of activities” [OUB 07]. For more details, see the IMS-LIP model and its questioning about interoperability with standards [OUB 07].

Figure 2.13. Activity model for the ODL global process [OUB 07]

2.4.3. Adaptive models In section 2.2.3.5, we explained how adaptive models address the need to personalize the digital learning environment. We recalled in this regard that this customization cannot be done without a good knowledge of the device. We must indeed be able to describe pedagogical paths that respond precisely and dynamically

Modeling Instrumented Activity at the Heart of the Virtual Environment

101

to the expectations identified by each learner. But make no mistake, these are not prescribed routes but adapted to everyone’s needs. “If the environment is well designed, the learner can navigate at will. It is not a question of making program-courses, but of guiding courses that offer the possibility of advancing in a non-linear way, where benchmarks are provided rather than stages. To consider the learner as a learner involves considering him as a person in his own right, who has the opportunity to take a path that suits him, knowing that those which are offered to him are conceived from a learning perspective. So, there is no good or bad course but a path that makes sense…” [SIM 10, paragraph 12]. The challenge is to make the system as responsive as possible to each learner’s situation, while preserving the pedagogical practices chosen by each teacher. It is for this reason that the model is said to be adaptive, because the system must dynamically adapt the educational pathways according to the expectations of each other. In 2011, the personalization was the subject of a French-speaking workshop at the EIAH 2011 conference. The results of this workshop were published in the form of a review article in the STICEF journal [LEF 12]. Since 2011, the question of personalization in the context of computerized training has been addressed by the international community as part of the PALE workshops. Traditionally, adaptive models are built from ontologies as part of instructional engineering [BEH 09]. Indeed, to “facilitate the research, adaptation and composition of services, it is essential to have a semantically rich description of services. Ontologies are now widely used to meet this need” [ZNI 12, p. 20]. Significant progress has been made in personalizing learning. Bejaoui, Paquette, Basque and Henri [BEJ 15, p. 4] are currently building MOOC scenario adaptive models, which support personalized learning based on these ontologies (see Figure 2.14). “[They build on] advances in personalization of learning and learner assistance in digital learning environments as well as the contribution of ontological modeling to these technopedagogical developments” [BEJ 15, p. 2].

Figure 2.14. Higher level of the ontology of a MOOC scenario supporting personalized learning [BEJ 15, p. 4]. For a color version of this figure, see www.iste.co.uk/trestini/learning.zip

102 Modeling of Next Generation Digital Learning Environments

Modeling Instrumented Activity at the Heart of the Virtual Environment

103

2.4.4. Systemic models of activity As we said in the introduction and at the beginning of this chapter, each VLE generation [BAR 94] corresponds to forms of mediations and models to represent them. The forms of mediation have been mainly described and studied in the framework of activity theories. Table 2.1 lists these forms of mediation and activity levels for the first three generations. Generation

Mediation

Level of activities

First

Individual mediation [VYG 97]

Mediation of activity by artifacts

Second

Collective mediation [LÉO 78, BLA 95, BLA 99, MCD 99, JAR 03]

Mediation system spread across a collective level: activity system

Third

Intercollective mediation [ENG 00, ENG 05]

Mediation system spread across an intercollective level: interdependent activity systems

Table 2.1. Different generations of activity theory, according to Engeström [ENG 00, ENG 05] cited by Laguecir et al. [LAG 10, p. 49]

Each of them corresponds to a model that we will present briefly. However, we will have the opportunity to review them in detail in Chapter 3. – The first model focuses on the individual level of the mediation of the activity by artifacts (tools). Vygotsky argued that the “Stimulus–Response” (SR) relationship between a human subject and an object is never direct. X represents the role of object mediation in the activity. For example, Vygotsky considers that consciousness emerges from human activity mediated by artifacts (tools) and signs.

Figure 2.15. The structure of mediated activity [VYG 78, p. 40]

104

Modeling of Next Generation Digital Learning Environments

– The second model24 is not strictly speaking a visual model. Based on the work of Léontiev [LÉO 76], it highlights the collective nature of the activity. “The individual is placed in a community, which corresponds to an extended system of mediation at the collective level: the system of activity” [LAG 10, p. 50]. Léontiev [LÉO 76] proposed to distinguish three levels within the activities. We have explained them in section 1.2.5.2: operations, actions and activities. We recall only that the operations constitute the basis of the activity and correspond to actions whose production has been automated by dint of successive achievements. Note that an action can be at the level of an activity or an activity can be an action in a process of more general scope [BEA 10]. – The third model is that of Engeström’s [ENG 87], which expands the previous triangular model by adding the subject within the community to which it belongs. Indeed, this model “does not allow us to consider all the relations that exist between an individual and the rest of his environment, in particular […] the fact that the activity is a collective phenomenon” [BOU 00, p. 48]. Engeström brings to the previous model the mediatization of two other relations: the first by more or less explicit rules between the subject and the community; and the second by a division of labor between the community and the object. The model in Figure 2.17 is the most recent representation. “[It] is finally one of the last major evolutions of this theoretical framework: the basic structure of an activity is in a bitriangular form whose angles identify the various parameters of human activities, thus taking into account the social dimension or activities as well as the mediatization of actions by the tools mobilized in the contexts observed” [BEA 10].

Figure 2.16. Individualized relationship mediated after Kuutti [KUU 96] quoted by Lewis [LEW 98] 24 If we ignore the model proposed by Kuutti [KUU 96] quoted by Lewis [LEW 98], in Figure 2.16.

Modeling Instrumented Activity at the Heart of the Virtual Environment

105

Figure 2.17. Engeström’s model [ENG 87]

What Engeström also says is that the system of activity evolves and draws its strength from the succession of resolutions of internal contradictions appearing within the system of activity. It is then necessary to distinguish four degrees of possible contradictions. The contradictions are: – primary, within each component of the central activity; – secondary, between the constituents of the activity system; – tertiary, between the central motive of the activity and the motives of other “culturally more advanced” forms of activities which may tend to replace the central motive of the activity; – quaternaries, belonging to networks of similar activities, between the central activity and neighboring or peripheral activities. As Bruillard points out [BRU 04, p. 3], this model can serve as a framework for analyzing human activity, but it also gives elements “showing how an imbalance, often a set of contradictions in the different intervening elements and a constraint at a given moment (often with a time limit), leads to transcending the initial framework to define a new system of activity, constituting a high form of learning”. This is what Engeström [ENG 87] calls expansive learning. The act of learning is part of the transformation of the activity system. According to this theory, learning is not considered here as a mere accumulation of information, but as the reorganization of a system of activity.

106

Modeling of Next Generation Digital Learning Environments

“This theory is so successful in the CSCW25 only in design issues” [BRU 04, p. 4], as we have seen in the example of the design of a rudder in section 2.2.2.2 or in the article by Béguin and Cerf [BEG 04]. Thus, among the ENA design models, systemic models derived from theories of activity not only facilitate the analysis of these environments, but also conceptually equip design processes, even if, at the same time, developments seem so necessary [BEG 01, p. 2]. Nevertheless, these last authors qualify their statement by adding that “the theories of the activity constitute a field of research more immediately usable for psychology (and in particular educational psychology) than for design sciences…” [BEG 01, p. 2]. To end this section, let us note that the Engeström model [ENG 87] has the unquestionable advantage of integrating the individual into his community, which corresponds to a vision of the extended mediation system at the collective level. It is more difficult, however, for us to understand how it represents an intercollective mediation as the author stated, quoted by Laguecir et al. [LAG 10, p. 49] in Table 2.1. From our point of view, a mediation system intervening at the intercollective level, in other words a network of interacting activity systems, would be better represented by the model that we propose a little later in the book (section 3.4.2, Figure 3.11). This point of view obviously does not detract from the incredible advances Engeström’s model has made in the analysis of activity systems. On the contrary, it paves the way for a whole range of possibilities in which we would be happy to list26 our model (see Figure 3.11). We refer the reader to section 3.4.2 for more detail. 2.4.5. Systemic models of complexity What systemic models of complexity bring to analytical models is first and foremost relative determinism. By referring to Engeström’s expansive model of learning, Virkkunen [VIR 07, paragraph 16] recognized that the analysis of internal contradictions within a system of activity is a long and complicated process. He takes as an example where, in an organization, the change in one of the activity mediators has provoked new internal contradictions within each element and between the elements of the system of activity. He seems determined, however, to follow the standard model of Engeström’s system expansion process for analysis, not without difficulty, however. 25 Journal of Computer-support Cooperative Work (CSCW, vol. 11, nos 1–2, 2002) on activity theory and design, written by Nonnie Nard and David Redmiles. 26 At first only, because we will see that the model proposed at this stage will experience, during our development, another evolution.

Modeling Instrumented Activity at the Heart of the Virtual Environment

107

In this case, which serves only as an example, the upholders of the systemic modeling of complexity admit their inability to provide all the details of the mechanisms that led to the emergence of these contradictions; it is the same for those who have allowed their resolution. They will nevertheless be able to speculate about final states without being able to explain precisely by what path they have arrived. According to Rochet [ROC 13b], it is the result of the large number of states that a system can take, given the interrelations that its parts can build, that makes the system complex. These states are not predictable by deterministic analysis when the system is open. Many of these states are emerging. It is in this sense that we can consider, in a simplified way, that “systemic modeling of complexity extends and completes deterministic analysis by introducing the notion of emergence” [ROC 13a]. It is then necessary to proceed by constructing models that will represent the possible states of emerging phenomena. The systemic modeling of complexity is an art by which the modeler expresses his vision of reality. In this sense, it is a constructivist approach; just as two individuals may have different knowledge of the same knowledge, two modelers will not necessarily build the same model. To model would be to admit that one cannot know everything: “the way is built by walking”27 said the Spanish poet Antonio Machado. We will not develop here the description of this model because we will return to it very extensively in Chapters 4, 5 and 6.

27 “Se hace camino al andar” in Spanish. Poem by Antonio Machado, available at: https://es.wikisource.org/wiki/Proverbios_y_cantares_(Campos_de_Castilla); shortened and translated English version available at: http://www.ttischool.com/blog/poem-week-caminanteno-hay-camino-antonio-machados-translation/.

3 Models of Instrumented Activity Challenged by Technopedagogical Innovations

3.1. Introduction Our first research studies strived to highlight the benefits of modeling instrumented activity by using the design and analysis models that we have just reviewed (section 2.4 in Chapter 2). In this new chapter, we account for the limits of these models when confronted with changing external conditions. Indeed, the previously mentioned models hold themselves together thanks to their initial structure, their skeleton, created to respond to a well-identified need. However, they must also resist pressure put onto them by the many waves of technopedagogical innovations that have swept over the education and training scene. Subjected to such pressure, their structure can change shape or restructure itself to resist changes in context. We can now measure how crucial it is to study their resistance to changes in initial conditions. This resistance to change and the structural transformations that this resistance entails will be the subject of this chapter. A necessary digression about the foundations of the theory of complex systems (section 4.1 in Chapter 4) will help us more easily express our arguments in favor of applying this theory to the modeling of latest-generation virtual learning environments (section 4.2 in Chapter 4). The suggested transformations in this chapter, based on design and analysis models of instrumented activity, are those that we have explored in our preceding work. They have greatly evolved since then. The most recent evolutions, which emerged from the application of the theory of complex systems, are justified in Chapter 4 and explained in Chapter 5.

Modeling of Next Generation Digital Learning Environments: Complex Systems Theory, First Edition. Marc Trestini. © ISTE Ltd 2018. Published by ISTE Ltd and John Wiley & Sons, Inc.

110

Modeling of Next Generation Digital Learning Environments

The technopedagogical innovations that have swept over the education and training scene have been an unforgiving test for the models that claim to represent VLEs, because these innovations change the context in which the models are introduced. However, they are not the only parts that can modify this context: certain socio-economic and political changes can also permanently impact the systems by changing the rules and codes of e-learning, for example. This context can also be referred to as “epistemic”, as it is characterized by a set of circumstances that undeniably interact with the modeling action. The epistemic context refers to a collection of data that were seen as reasons to believe in these successive technopedagogical innovations and thus shaped the actor’s beliefs [OUZ 11]. Piaget introduced the notion of an epistemic framework to design “the framework of meanings in which society inserts objects and events and which has an influence on the way in which we assimilate and interpret every particular experience. Therefore, the epistemic framework refers to the social system of meanings which partly affects the way that we interpret the experience”1. Moreover, if we use these models, it is indeed to describe and understand human behavior and the different roles played by digital techniques in the field of learning. Therefore, we should not neglect to take the epistemic context into account. Models are also made to evaluate the quality of VLE design as well as VLE potential in terms of its targeted audience, devices and technopedagogical choices. A model is characterized by its strong dependence on context. This context is itself characterized by a very large variety of uses and by permanent changes on a technological and pedagogical level (technologies that are increasingly mobile, distributed, collaborative, interactive, adaptative and easily accessible). How can current models resist or adapt to the multitude of these new contexts and problematic uses related to the continuous and rapid evolution of technopedagogical innovations? To answer this question, we must also consider that “behind these uses, several different processes are at work, and it is not enough to make decisions solely on their existence and frequency. It is the analysis of a ‘complex system’ and its evolution which is required” [BRU 06, p. 283, cited by POY 12, paragraph 5]. What models can be built which are able to analyze, compare and design these systems? How do these models enable the design or adaptation of criteria and analysis factors when technologies are distributed, adaptive, collaborative and interactive? How do they observe and analyze the coevolution of humans and the technical devices that operate in these contexts? To come up with some answers to these, we will set out some concrete examples, which are mainly based on personal observations and our past research. Each time that we make reference to these, we will systematically cite them. 1 Source: http://www.fondationjeanpiaget.ch/fjp/site/ModuleFJP001/index_gen_page.php? IDPAGE=323&IDMODULE=72.

Models of Instrumented Activity Challenged by Technopedagogical Innovations

111

3.2. The Vygotskien model and its expansion The purpose here is to test Vygotsky’s first triadic model [VYG 78, p. 40], which is used to analyze the use of technological objects in more traditional school or university contexts. Vygotsky’s model essentially tells us that knowledge emerges from human activity, which is mediated by artifacts (tools) and signs. The triadic representation that he proposes (Figure 3.1) perfectly reflects this idea.

Figure 3.1. First-generation model of mediated action – Vygotsky

However, is this representation now enough to inform the modeler of all the phenomena likely to emerge throughout the introduction of the latest generation technological “object”? Moreover, is it enough to accompany and support its reflexive pathway (see the function of a model in section 2.1 in Chapter 2)? If not, which other model can we use, or how can we modify this one, so that it facilitates observation of new or unexpected phenomena? 3.2.1. Digital ink: towards a new virtual learning environment design To begin to answer these previous questions, let us take a particularly simple example: we suggest that a lecturer conducts a lesson in a lecture theater using a computer that can display the content of the lesson. In this model, the teacher is represented as a “subject”, who uses a “tool” (the computer) with the aim of carrying out a lesson: the “object”. In this example, no collaborative work and no interactive activities are carried out during the lesson. At first sight, Vygotsky’s model is very much sufficient to represent the context of the study and make an analysis. However, experience shows that this particular situation can be seen by the teacher in two very different ways: the tool can either be unconditionally adopted or be systematically rejected. In the first case, the artifact is used, and in the other, it is not used. Thus, what could be the reasons for such different behaviors? Which model will allow us to determine them? For the first question, we can say that there are numerous

112

Modeling of Next Generation Digital Learning Environments

reasons linked specifically to the actor’s perceptions of the utility and usability of this tool. However, another element seems to come from the nature of the content to be taught (whether it originates from hard sciences or social sciences, for example), content which is furthermore absolutely not represented in Vygotsky’s model. In other words, the teacher’s use or rejection of technology could also be considered a “didactic” problem, provided all other factors remain equal. To understand the reasons for this and to show how the model’s consideration of all potential causes of non-usage (whether they are didactic, pedagogical or technical) is fundamental, let us now describe in detail the context of a more in-depth study [TRE 10b]. The aforementioned study focuses on the device comprised of a Tablet-PC (different from a Tablet) associated with an online textbook [TRE 10b]. The Tablet-PC (see Figure 3.2) makes it possible to project what the teacher previously used to write on a board, while the online textbook gives online access to all the pedagogical content (videos, soundtracks, texts, etc.). With a simple click of the mouse, the students can then listen and watch at home as many times as needed soundtracks and course videos, known as podcasts. As a concept, podcasting is an undeniable benefit for university teaching [KAR 01, LEB 02]. It contributes to the enrichment of face-to-face courses by offering the student the possibility to return to it and consult some of the course again. Nevertheless, the didactic quality of these podcasts still seems questionable to us. We will not call into question the pedagogical benefit of the podcasting concept here, but rather look at the didactic quality of its content. Traditionally, a video-lesson is a true copy of the lesson in the lecture theater; in other words, a perfect reproduction of the teacher’s actions (and possibly of a slideshow or a computer-assisted presentation) on a visual channel, along with explanations on an audio channel. “It is quite well suited to lectures in the social sciences where speech is paramount, this device appears much less successful in natural sciences that rely primarily on the teacher’s written records and not only on his speech” [TRE 10a, TRE 10b, paragraph 1]. It is true that “the audio and video trail showing the construction of a schema by a teacher often has more value than the schema itself” [POR 10]. Yet, the recreation of these written traces through a traditional video course has proved difficult, even impossible. To offset this difficulty, we have proposed an original video-course device named “digital ink” [TRE 10a, TRE 10b]. It allows us, thanks to a Tablet-PC, to project a collection of resources that the teacher chooses to use on a large screen. These include not only course slides, illustrations, videos and explanatory diagrams, but also the teacher’s written traces.

Models of Instrumented Activity Challenged by Technopedagogical Innovations

113

Throughout the whole course, all the resources (including written traces) used by the teacher are recorded in a synchronous manner with his own comments. After a breakdown of the sessions and sequences of the video-course, these are then broken down into parts (chapters and sections) in the tree of an online textbook for students. The latter can then be used to consult all or part of the course at the student’s will.

Figure 3.2. Tablet-PC tool (Surface Pro)

Figure 3.3. Sample writing course on Tablet-PC [TRE 10b]

Our experience shows that if the tool is readily used by teachers in the field of hard science, on the other hand, it is rejected by teachers in the social sciences field and this for the same configuration of teaching-learning or for the same teaching position. Conversely, during traditional video-courses or when the teacher is simply

114

Modeling of Next Generation Digital Learning Environments

filmed as they carry out their lesson, there are a lot more followers in the social sciences than in the hard sciences. The content to be shared triggers this significant difference in usage. From this experience, we came to the conclusion that, without any means of distinguishing the didactic artifact (the type of course content to be transmitted) from the technical and pedagogical artifact, all three buried in the heart of the “tools” clusters of Vygotsky’s model, we probably would not have been alerted by the existence of a strong link of dependence between the use of the Tablet-PC and the didactic content to be transmitted, and this for the same pedagogical configuration (scripting). In other words, Vygotsky’s model was not conducive to the detection of a possible “instrumental conflict” [MAR 11] between at least two of these three artifacts. We therefore risk proposing “an engineering approach which is too quick or incomplete” [MAR 11, paragraph 4]. By instrumental conflict, we mean “the consequences of an interference which can emerge between one or several artifacts involved in the situation” [MAR 05, p. 387]. In fact, we find it useful to propose an expansion on Vygotsky’s model. This should be able to reveal the three artifacts buried in the “tools” clusters (see Figure 3.4).

Figure 3.4. Expansion of Vygotsky’s triadic model [TRE 10a]

3.2.2. Use of tablets (iPads) in school contexts Furthermore, to show the limits of Vygotsky’s model of the structure of mediated activity (Figures 3.7 or 3.8), and further recognize the benefit of having

Models of Instrumented Activity Challenged by Technopedagogical Innovations

115

passed from this model to that of Engeström [ENG 87], let us quote another example. This deals with the use of the “tablet” (different from a Tablet-PC) in a school context. The “tablet” has been the subject of a comprehensive national equipment plan2 as part of the “digital plan” gradually rolled out in France from the start of the 2015 academic year.

Figure 3.5. School equipment plan with tablets (source: French government site, 12 May 2016)

The incompleteness seen in Vygotsky’s model is due to the fact that this technological innovation completely changes the relationship between the “subject” and the “object”. In fact, a study by Taylor, 2006, cited in Sharples [SHA 07, p. 9] shows that by redefining the learning contexts, these virtual spaces change the student’s relationship with knowledge: “the mobile environment is eminently suited to supporting learning outside the context of curricula, institutions and timetables” In the context in which it is used, the tool “tablet” (iPad) would no longer benefit the student’s production activity (like the classroom computer did) but would benefit the sharing and intake of content in a collaborative and interactive way [MEL 10, p. 1], at least with the uses observed. The structure of Vygotsky’s model is therefore disrupted, and its application compromised, as it does not allow us to represent this new data. By changing the learning context, tablets are therefore considered as “‘consumer’ and entertainment tools, according to dominant culture, and this leads us further away from the meaning that the student gave to the tablet, that is, ‘a creation and learning tool’” [MAR 13]. We could then discover contradictions in the students’ perception of the utility of the tool [MAR 13]. 2 At least 1,256 schools and 1,510 French high schools were equipped with tablets in September 2016. That is nearly a quarter of colleges that have joined the digital plan. More than 175,000 students will be equipped with digital tablets, co-financed by the state and local authorities; information dated 12 May 2016, French Government website, available at: http://www. gouvernement.fr/action/l-ecole-numerique.

116

Modeling of Next Generation Digital Learning Environments

To provide evidence of these contradictions, Martineau and Barma [MAR 13] have chosen to use Engeström’s model [ENG 87], which is itself a transformation of Vygotsky’s model. Considered as an expansion of the latter, Engeström’s model not only takes into account the student’s “production” activity, but also the “intake” of information (Figure 3.6). Furthermore, it allows the analysis of the contradictions at work between these two activities within each cluster (primary contradictions) or between these clusters (secondary contradictions).

Figure 3.6. The representation of activity systems theory according to Engeström [ENG 87]

Note that “Yrjö Engeström considered Vygotsky’s triangle narrow and inadequate to describe human activity and all factors affecting on it”3 (Lönnberg, Paloheimo, Lahti, Keltikangas and Mannilan). It is therefore with the same arguments and purpose that he created this model that numerous researchers have used thereafter (such as Martineau and Barma [MAR 13], taken here as an example). 3.3. Expansion of Engeström’s model 3.3.1. Looking for a unifying model in long-distance learning Throughout our first research studies ([TRE 05, BEN 06, TRE 10a] focusing on the analysis and design of a VLE – mainly as part of EAD), our intention was to roll out a flexible conceptual framework that can be relevant to describe not only the diversity 3 From Lönnberg J., Paloheimo A., Lahti L., Keltikangas K. and Mannila L. (n.d.). Methods and Results in Computing Education Research. Available at: http://www.cs.hut.fi/u/aura/ paivin_seminaari/activity theory.pdf.

Models of Instrumented Activity Challenged by Technopedagogical Innovations

117

of the actors and the tools involved, but also the variety of methodological and pedagogical approaches used as part of the EAD framework. It must also facilitate the study and coordination of different tasks and different actors involved in this type of environment, all while supporting a theoretical framework accepting the notion of collaborative learning. To that end, we have often used systemic modeling by Engeström [ENG 87] derived from activity theories (Vygotsky, Léontiev). As we have seen in section 2.4.4 in Chapter 2, this model allows us to concretely represent this conceptual framework, and to exploit the different relationships between the clusters of its internal environment (tool, subject and object) and those of its external environment (the community, the rules and the division of labor). Figures 3.7 and 3.8 illustrate these two environments, respectively.

Figure 3.7. Individual relationship mediated through Vygotsky’s model [TRE 10a]

More specifically, this model allows us to plan a design process or to analyze a VLE by exploring each of the triads that define relationships (materialized by lines). Pressures or contradictions that may follow (or those that already exist) can then be detected thanks to this model (Figure 3.8). For example, the way in which an object is shared with the community via a tool can be studied. Possible pressures between these clusters may then be revealed or even avoided by analyzing this particular triad (in orange or light gray in Figure 3.8).

118

Modeling of Next Generation Digital Learning Environments

Figure 3.8. Basic structure of an activity according to Engeström’s model [TRE 10a]. For a color version of this figure, see www.iste.co.uk/trestini/learning.zip

It is exactly through several attempts to design or analyze long-distance learning devices founded on constructivist learning environment designs that we have found the first limits of Engeström’s model and our intention to propose an expansion of it4. We felt the need to break down each of the clusters (or components) of the model in order to allow us to reveal the existence of certain system components, “hidden” or “buried” inside each of the model’s clusters. For numerous reasons that we will not touch on here (analysis of instrumental conflicts, actors of different profiles, communities of different actors, etc.), we had to actively open these boxes (the model’s clusters) in order to reveal the presumed components. With these components, we then had to pursue our analysis a little further. We then proceeded to break down each of the model’s poles into clusters by establishing the new relationships between them [TRE 10a]. After the first representation proving this intention (see Figure 3.9), a model highlighting successive interlockings emerged. In fact, each pole could sub-divide, but each sub-pole could also sub-divide. The unexpected homology5 with Mandelbrot’s fractal theory [MAN 09] (see Figure 4.1) has led us to borrow vocabulary from this theory and to label this model as self-similar, that is, having the same aspect at any scale (it is a form of expanding symmetry). We will therefore leave this observation for now to avoid arguing too much about a possible similarity with this theory, which asserts that an order would exist in what could be called a “systemic disorder”. 4 This expansion has nothing to do with the process of expanding the activity system [ENG 87] cited in section 2.4.5 in Chapter 2. 5 In evolutionary biology, homology refers to the same characteristic observed in two different species (source: Wikipedia).

Models of Instrumented Activity Challenged by Technopedagogical Innovations

Figure 3.9. Modeling from system analysis data [TRE 05]

119

120

Modeling of Next Generation Digital Learning Environments

This expansion of Engeström’s model is therefore created slowly and at different stages. Furthermore, each stage leads to the expansion of the model in a new spatial dimension: first in two dimensions in a somewhat disordered way (Figure 3.9) [TRE 05], then, still in two dimensions but this time with more regularity (Figure 3.10) during the design of a CLE device [TRE 10a, TRE 10b], and finally, in three dimensions (Figure 3.11), during the analysis of the implementation of an online textbook within PLACE6 (Lorraine educational DWS) [TRE 09a]. Figure 3.9 above truly shows our first “intention” to expand the model as we envisaged in 2005. 3.3.2. First expansion of Engeström’s model for designing a CLE Concerning the representation that follows, several contextual elements would be useful for understanding. We were looking [TRE 10a] to outline the designer’s reflexive course for the design of virtual learning environments based on constructivist learning environment designs (CLE7). Essentially, we proposed an approach based on activity theory (Vygotsky, Léontiev) and its systemic representation (Engeström), which allowed us, after several adjustments to the basic model, to exploit the clusters of the internal environment of the targeted learning world (the tools, the subject and the object) and those of its external environment (the community, the rules and the division of labor). They then in turn became pole-cores around which the other clusters pivot by creating networks of relations. From these came the considerations accompanying the reflexive process carried out by the designers of CLE in order to model them finely and thus to facilitate their design, scripting and production [TRE 10a]. All this brought particular light to the CLE modeling process. The relationship networks of Engeström’s model were highlighted and allowed projections of other networks to be built. The approach we were describing then involved providing CLE designers with the means to work as hard as they could to investigate the existing relationships within the system and its classes. “By adding subclasses, each class develops in ‘clusters’ of components and subcomponents that are attached to each of the model’s clusters. Nested in each other like Russian dolls, each component or subcomponent contains smaller ones and is able to reveal them. They are thus assigned properties one by one, also called 6 PLACE: “Plateforme Lorraine d’accessibilité et de communication pour l’éducation” is the solution of digital workspaces (DWS) for high schools in Lorraine implemented in the years 2009 and 2010. 7 Constructivist Learning Environment.

Models of Instrumented Activity Challenged by Technopedagogical Innovations

121

attributes. Figure 3.10 illustrates this development in clusters in the particular case of the ‘tools’ cluster” [TRE 10a, p. 58].

Figure 3.10. Model with development of the “tools” pole

We therefore added finer components to the “tools” cluster (Figure 3.10): – pedagogical artifacts and the types of activities associated with them, that is, collaborative or cooperative activity, supportive activity, support for personal work, etc.; – didactic artifacts, that is, the teaching content to be transmitted that result from a didactic transposition of “scholarly knowledge” into “knowledge presented”; – technical artifacts (support, architecture, networks, etc.). These artifacts, whether pedagogical, didactic or technical, contribute to the development of the human activity of learners. We thought that these new components, revealed by explaining the structure of the “tools” pole, would give the designers a better readability of the relationships that the subjects are likely to

122

Modeling of Next Generation Digital Learning Environments

maintain with these artifacts. Possible “instumental conflicts”8 between these three artifacts could also be avoided by pointing out to designers the pitfalls to be aware of. Some predictions could be deduced, for example, the type of pedagogical scenario that would seem best suited to facilitate the appropriation of such teaching content, with this type of media. The object must be transformed by the subjects of the activity system. It must then maintain the modeler’s attention. The objects of the activity can be physical objects (for example, objects to build), software objects (for example, a computer program) or conceptual objects (for example, a theory, a theoretical model that is negotiated). All the contents envisaged in a cognitive aim become objects of learning, whether as a matter of conceptualization, instrumentation or curriculum. The subject remains the vital cluster in CLE design: the tools and the object are at its service. Relationship networks link it to other components and subcomponents that the model has. This occurs as relationships are defined between the speakers in view of the roles they are called upon to play: the subject becomes a learner, a teacher, a designer or a coordinator of the training and the tutor groups he organizes and the learning pace he adjusts to. Then communities and their intersections are created, which the subject fits into. In conclusion, we provided an overview of the reflexive path that should be taken if we took into account the clusters of the external environment, such as the cluster for the rules, the division of labor and the community. We did this by highlighting the transformations that would likely cause the model to become the target. To further explore Engeström’s model, we planned to target the secondary clusters, including those we recognized as part of the subject area, namely the division of labor, the community and the rules. Thus, and having regard for the relationships that bind the designer to the community and the division of labor, the former is led to establish rules beforehand that: – set the roles and responsibilities of the actors; – establish the distribution of tasks. These are shared and undertaken by the community, which is undeniably a major component of the activity system. After having carried out the process of designing a CLE from a continuous reflection made up of intentions, it was necessary to recognize that this occurred not in the world of action, but in that of ideas.

8 Marquet P., “Intérêt du concept de conflit instrumental pour la compréhension des usages des EIAH”, Proceedings of the EIAH Conference, 2005.

Models of Instrumented Activity Challenged by Technopedagogical Innovations

123

Designing CLEs is an approach that simultaneously aims to design, explore and systematically and dynamically exploit the components of a constructivist learning environment, in order to establish the idealized representation. Fields as different as tools, subjects and objects are brought together. In practice, the proposed approach advocated the use of modeled representations of CLE. It was a step prior to any scriptwriting and practical learning of environments conducive to learning. As in the design system [SIM 91], activating the process would result in an intentional action plan [LEM 08] related to the project. This should first of all serve to guide the work of the CLE designer, but also to help him to solve the problems encountered and to constantly adjust the interfaces between the real world and the represented world of the object under consideration. 3.4. New context of usage and new expansion The second major expansion of the Engeström model was discussed in the previous section. Showing too much resistance to the context of the study, the model was not developed in two dimensions as before, but this time in three dimensions (Figure 3.11). It was during a “study on the Environment PLACE textbook” that this modeling need appeared [CHE 10]. The proposed modeling was also presented at the International9 Cned-Eifad symposium in 2009 [TRE 09a]. It was subsequently published in the Distances et savoirs journal [CHE 10]. Let us now go back to some contextual elements before justifying the need for model expansion. 3.4.1. Digital workspaces in schools: online text books The use of a DWS in a school establishment could be questioned in terms of the organization and the institution of engineering, pedagogy, student learning or even the place, where parents follow up on their child/children’s progress. The research we were interested in here was concerned with these different aspects, focusing in particular on the use of a specific object, which is often integrated into school DWSs: the online textbook. “In addition to being innovative in the context of exploratory research, it had the heuristic advantage of being amenable to evaluation according to four perspectives (administration, teachers, students, 9 Cned-Eifad 2009, Futuroscope site: “Distances et Savoirs à sept ans !”.

124

Modeling of Next Generation Digital Learning Environments

parents) and to allowing the observation, characterization and qualification of mediations at work in the school space.” [CHE 10] The first objective was then to identify the functions of the textbook (operational framework) and to observe its first uses (usage framework) by the teachers, the pupils and their parents, as well as by the administration and the supervisory staff (inspectors in particular), in order to measure the effect of this tool on usual practices. The second objective was to characterize the practices related to this tool (institutional practices, pedagogical, learning, communication, monitoring of students, etc.) and highlight the continuities or disruptions caused by the usual practices. The observation of the real uses of this tool by the beneficiaries should also make it possible to compare them with those prescribed by the designers. 3.4.2. Second expansion of the model: facilitating the analysis of online textbooks The issue was mostly about understanding how the shift from access rights granted to a relatively small number of actors (within the class) to access rights granted to an expanded community (the educational community presented virtually on the Web: parents, students and administration) impacted the way teachers described and prescribed classroom instructional activities. To what extent and by what method was the training activity of textbooks reworked by putting them online? In what way and how was the pedagogical activity of the teacher as a whole reconfigured? It was these reconfigurations in particular that interested us and that we wanted to analyze. Once again and beyond the general problem raised, it is worth noting that it is by changing the rules of the analysis that we are testing the resistance of the model used (Engeström’s or its previous expansion) and that we are potentially showing the limits. For a homogeneous and well-identified community, the models mentioned above withstood the analysis relatively well. The subject cluster was formally recognized in this model as being an “individual” or “individuals” or “a group” to be analyzed. However, when it came to understanding the transition from a possible consultation by a community of students to a possible consultation within an enlarged community (the community of parents, students, teachers and even the administration), the model clearly showed its limits. Indeed, we have not yet found any reference to separate groups in Engeström’s work, although some authors [LAG 10, p. 49] place his model “... at an intercollective level. Evidence for this is the way in which ‘communities’ communicate with each other in this model by using a common ‘object’ (the textbook) which was not ‘representable’ using the previous models” [OUB 07].

Models of Instrumented Activity Challenged by Technopedagogical Innovations

125

A new way of developing the model was then proposed [TRE 09a] by taking the model out of its planar representation to occupy all the space this time (Figure 3.11). This representation allowed us, for example, to study the way in which the object “textbook” was shared through the PLACE tool within each community. However, it also became possible to analyze certain pressures by sharing this object between the different communities. This is the benefit of this new expansion of the model, which allows this representation and thus facilitates the analysis. We could continue to show the limits of this model by using it, for example, in the more general context of analyzing the uses of school DWSs (and no longer only those of the textbook, which constitutes only one of its software bricks). It would show very quickly its inability to represent “separately the uses of different functionalities that do not all behave in the same way even though part of the same global entity, the DWS” [POY 12, paragraph 12]. This dissociation has also enabled the latter authors to establish by other means “comparisons of uses or non-uses between these functionalities” [POY 12, paragraph 12].

Figure 3.11. Relationships between community activities

3.5. Expansion of pedagogical and training engineering models In our “overview” of models and modeling languages (section 2.4 in Chapter 2), we have presented some of the current models of pedagogical

126

Modeling of Next Generation Digital Learning Environments

(section 2.4.1 in Chapter 2) and training engineering (section 2.4.2 in Chapter 2). We have explained how these models have progressed in order to take into account, on the one hand, the evolution of teaching designs (active and collaborative pedagogies in particular) and, on the other hand, the evolution of the activities as a whole (and not only those of an educational or didactic nature) “within a global process of open and distance learning” [GAR 03]. Let us examine here how these models have resisted the pressure of these two constraints put in perspective and conclusively consider some possibilities of the global evolution of VLE models facing the emergence of new training methods on the Web, such as MOOCs by example. 3.5.1. Resistance to pedagogical engineering models Since the expansion of virtual learning environments, a major effort has been made to develop models and standards for their design. As pointed out by Pernin and Lejeune [PER 04b, PER 03], the initial dominant model focused on “a transmissive learning mode and behavioral theories of learning”. But again, learning is not simply a transfer of knowledge from resources made available to the learner. “The acquisition of knowledge and know-how comes from many sources and forms of varied activities [KOP 01] such as problem solving, interaction with real tools and collaboration with other actors. To improve and diversify online training, it has therefore become necessary to pay attention to activities in the learning process, in particular those of the learner, and to describe and organize these activities” [OUB 07]. From this observation, the first “content-centered” models did not resist the constructivist learning methods. This is why Koper [KOP 01] was at the heart of IMS-LD “process-centered” modeling. This approach “constitutes a real disruption by formally dissociating activity and resources and by clarifying the semantic relationships linking them” [PER 04a]. Since then, several researchers have been involved in projects to enrich these pedagogical models, notably by modeling collaborative pedagogical scenarios [PER 04b, FER 05] and by seeking to support the development and reuse of pedagogical scenarios [FER 05, LAF 05a, LAF 05b, OUB 07, VIL 07, etc.] with elements specific to each research project.

Models of Instrumented Activity Challenged by Technopedagogical Innovations

127

In particular, the desire to model the collaborative teaching scenarios marks a new break from IMS-LD models, or rather the appeal to expand these models. To illustrate this need, Ferraris et al. [FER 05] gave a concrete example of LOG (Lycée ouvert de Grenoble) teachers, which offers distance learning courses for specific audiences (sick children, high-level athletes on the move). “The problem that arises is the difficulty of planning collaborative activities for a virtual classroom beforehand, because the available environments only offer the teacher a series of dialog boxes and do not keep an explicit and reusable version of the actions carried out for this implementation. The collaborative scriptwriting expertise of teachers is therefore not capitalized” [FER 05]. Considering the pedagogical scenario created by the teacher as largely improvised and very evolutionary, these researchers wanted to integrate a dimension of regulation and observation into the model, in other words the possibility of observing what happens when the scenario takes place and allowing the teacher to intervene and modify it dynamically during execution of the task. Faced with a concrete pedagogical situation, they first asked themselves if IMS-LD could help them to formalize and model the proposed collaborative scenario as they pleased. They deduced that certain conceptual objects lent themselves quite well to IMS-LD modeling, while for others, the model was more resistant (problems of expressiveness in collaborative activities, difficulty in expressing the results of an activity, impossibility to regulate, etc.). Santos et al. [SAN 04] clearly announce that IMS-LD does not provide a specific descriptor to describe the interactions of collaborative activity members. They provisionally responded to this problem by defining as many individual roles as necessary, each assigned to specific tasks in parallel with the collective activity [FER 05]. They also decided to start from a “Participation Model” [MAR 04] that had been formalized in another context (that of groupware). Its purpose was to describe a group activity (collaborative activity), in which the actors are considered as active participants who build their own workspace and define their own working conditions, the commitment they make and the roles that they will play, all this in a logic of co-construction of the activity. But make no mistake, this activity obviously does not exclude the possibility of self-regulation, that is to say, that the concepts put in place allow us at any time to change the educational scenario. The scenario described in a free text is translated into the metalanguage IMS-LD and its expression is enriched by the use of the “metamodel of participation” [MAR 04].

128

Modeling of Next Generation Digital Learning Environments

Technically, the pedagogical scenarios are described by a set of descriptive metadata from IMS-LD (which target audience, pedagogical objectives). However, other concepts from the “participation model” enrich the collaborative model, such as the concepts of enclosure (place where the activity takes place), position (how the actors are positioned in this collaboration), roles, rules, etc. It is worth noting here that to model these collaborative learning scenarios, this team of researchers has defined a new language, the Learning Design Language, using the formality of the UML activity diagram. This remark is useful for understanding the choices we will make later. An example of a scenario description is presented in Box 3.1 and the corresponding activity diagram in Figure 3.12. The Committee reconvenes again to decide whether or not to publish The Fairy Gunmother (Daniel Pennac). Many members agree, but still a hand rises. It is a young woman: she objects: "I find it shocking that Mo the Mossi and Simon kill Risson by overdose. It’s not up to them to mete out justice!". It refers to chapter 32. A murmur runs through the assembly. The reactions are not long in coming. And you, what do you think? But the director looks at the time and proposes to continue the discussion at the neighborhood café. Goal Outline one’s position in a debate. Method Students learn about the instructions, then the principal appoints a reporter from among the students in the group and invites them to join the discussion forum called "debate space". In the forum, everyone will have to intervene at least twice, either to take a position and to advance a new argument on the actions of Mo and Simon, or to respond to an argument. After 20 minutes of discussion, the reporter will report on the main positions to the rest of the group, and after negotiation, will write a summary document. The summary document will then be presented to the director by the reporter.

Box 3.1. Descriptive text of the scenario [FER 05, p. 6]

Note that the choice of the UML language is very often mentioned when it is necessary to integrate other conceptual objects into the possibilities of EML languages (like IMS-LD). It is a modeling language that, upstream of EML-type languages, perfectly meets the need to model complex systems in general. More generally, it will be noted here that the interest of applying the modeling languages of complex systems (such as UML languages) to the design of ENA becomes progressively more precise with the technopedagogical evolutions of which they are the target. Moreover, the IMS Global Learning Consortium, the largest consortium

Models of Instrumented Activity Challenged by Technopedagogical Innovations

129

for the specification of educational content, is a method of computer development associated with or derived from the Unified Modeling Language (UML) [HEN 07, p. 19]. Many other experiences testify, on the one hand, to the difficulty that models have in resisting the technopedagogical waves experienced by the world of e-learning, and on the other hand, to the increasing use of modeling languages specific to complex systems (such as UML) for the design of VLEs. We cannot, of course, quote them all here; let us recall only by way of example that Lafourcade [LAF 05] proposed an approach based on model transformation for the design of ITS and, in particular, for the design of problem-solving situations on distance learning platforms. “The study of the different languages corresponding in part to this objective made it possible to highlight (i) ‘the current lack of models and languages for the stage of initial expression of needs’, as well as (ii) ‘the necessity to use UML to describe abstract models (analysis step) serving as a basis for the formal specification made by EMLs like IMS-LD’ (detailed design step)” [LAF 05, p. 215].

Figure 3.12. Diagram of UML activity scenario [FER 05, p. 6]

130

Modeling of Next Generation Digital Learning Environments

This led to the adoption of “a graphical modeling language that specializes in UML specifically for the design of problem-solving situations upstream of EML-type languages: it then covers the design phase for the initial expression of needs stages, analysis and design” [LAF 05, p. 215]. Subsequently, the modeling of pedagogical scenarios has been the subject of new developments, which we will discuss in detail in Chapter 4. One option considered (which still remains relevant) is to gradually integrate them into a more general business model representing this time the training device in all its dimensions, namely educational, didactic, technical, but also administrative, etc. “The pedagogical scenario leads to the creation of the training device insofar as it prescribes certain activities, proposes certain uses of the environment and the media, and generates particular mediations” [HEN 07, p. 21]. 3.5.2. Evolution towards training engineering models During the presentation of pedagogic engineering models, we had come to the idea that a too partial representation of the activity, focused, for example, on the pedagogical activities, suffered from a significant lack of representativity concerning the interactions between different types of activity. For example, educational activities could include relationships with other forms of activities, such as administrative, technical or training follow-up activities. Integrating all these activities into a more comprehensive VLE model was then seen as a priority; the goal is to build a business model that better takes into account the specificity of these management activities, even if the model becomes more complex. The management of this complexity, which is precisely the main object of this book, will be studied in Chapter 5. The model (or metamodel) to be built can then be considered as an expansion of the classical models of educational scenario because it takes into account, by integrating within it, other activities than those that traditionally revolve around the learner and his tutor. The increasing impact of various related factors (management, governance, etc.) on training activity is likely to be behind this development. This is probably also the result of the diversification and massification of the population attending open and distance training, which also suggests other developments with what the MOOC now allows. This therefore implies taking into account related activities such as the management of training plans, personalized training courses and associations between tutors and learners according to their respective profiles (skills, status and associated rights, or learner follow-up).

Models of Instrumented Activity Challenged by Technopedagogical Innovations

131

In this perspective, Oubahssi and Grandbastien [OUB 07] proposed a diagram, 10 which links and takes into account all of these activities (Figure 2.12). We have detailed it in the previous chapter. This model gives a global view of the activity in the process of open and distance learning. This led them to define a new concept: the “work unit”. 3.6. MOOC models to build In the beginning, French MOOCs were presented by the MENESR 11 “as an experiment that will evolve with use according to the users’ opinions”. It is precisely on this experimental and evolving aspect of the MOOCs that an important research work was conducted during the years 2015 and 2016 [CHE 15a, CHE 15b, TRE 15a, TRE 16a]. It aimed to understand what motivated their development in France and how models (mainly pedagogical) have evolved. The question of the social appropriation of this technopedagogical innovation was therefore central to study this evolution. It was first approached from the point of view of the representations of the professionals of online education [TRE 15a] and was then supplemented by that of the registered users [CHE 15a, CHE 15b]. In general, representations lead people to take a stand on topical issues, professional practices, social events or even technical innovations, in relation to their social background and their social and professional contexts. This is the organizing principle of social representations developed by Doise [DOI 86]. When the group is formed for professional purposes, the opinions and beliefs that are built within it are part of the so-called professional representations, which are “social representations of objects belonging to a specific professional environment and shared by members of the same profession” [PIA 99]. According to Abric [ABR 94], there exist within the social representations and by extension within the professional representations [BAT 00] central elements and peripheral elements. Central elements are defined as consensual, generative and organizers of the rest of the representation and “relatively independent of the immediate context” [ABR 94, p. 28, cited by LHE 10, p. 316]. They determine the general direction of the group’s relationship with the object, independently and beyond situational contingencies. By contrast, the peripheral elements may be subject to greater inter-individual divergences and are dependent on the immediate context [ABR 94].

10 Diagram of classes under UML (Unified Modeling Language). 11 MENESR: Ministère (français) de l’Éducation nationale, de l’Enseignement supérieur et de la Recherche (French Ministry of National Education, Higher Education and Research).

132

Modeling of Next Generation Digital Learning Environments

We cannot detail here all the results that this experimentation has allowed us to discover and instead refer the interested reader to the publications that discuss the subject. Note that at the beginning of the French adventure12, the peripheral elements of these representations, whether social or more specifically professional, showed the eagerness of French actors involved in MOOCs to move as quickly as possible the MOOC models that they considered to be rather confusing and globally unsuited to the French sociocultural context13. Considering that the French MOOC models were only a replica (or exact copy) of North American models that were at the least very different from one other (cMOOC, xMOOC, hybrid, etc.), the main difficulty mentioned was that there was no common representation upon which it was possible to rely. In the minds of online education professionals, the French MOOC model would have been built only through the image conveyed on social networks and without any apparent desire to reshuffle14. In fact, it seemed urgent that France first adopt an educational and economic MOOC model (in other words, a “theoretical” model) before considering the conceptualization and formalization of an “object” model for the design and development of French MOOC analysis. We are currently observing that a form of penchant for xMOOCs (as opposed to cMOOCs) is developing in France and that the formalization of an economic model is becoming clear with the institutionalization of this innovation through the creation of a public interest group called “FUN MOOC” [TRE 16a, p. 45]. As for the “object” models that draw our attention, besides some recommendations for their design (for example, the MOOC named “Créer son MOOC de A à Z” by Mathieu Cisel) and the application of models coming from distance learning, it seems that everything still needs to be built. The purpose of our current research is precisely to help fill this gap. 3.7. Conclusion on the resistance conditions of virtual learning environment models By taking on a role as designer or analyst in turn, these few examples allowed us to illustrate the way in which the models are modified to respond to a change in socio-technical context and in particular when introducing technopedagogical innovations. We will then study other cases, especially when the decision to change 12 Our experimentation started at the very beginning of the French experiment, that is to say from 30 January to 20 February 2014, which corresponds to about the end of ITyPA season 2 and the first month of operation of FUN. 13 Even if they are well aware that, in essence, their influence is international. 14 Twenty-two percentage of professionals surveyed thought that the pedagogical model underpinned by MOOCs consisted of a mix of transmissive (36%), connectivist (45%) and constructivist (23%) conceptions of learning!

Models of Instrumented Activity Challenged by Technopedagogical Innovations

133

or evolve the model will be due to rapid or large-scale systemic changes. This is the case, for example, of the MOOCs that we have mentioned and which will be taken up and developed in detail later. This is also the case of the phenomena of rationalization of distance education, or more broadly of its industrialization [TRE 14], the effects of which have already been indicated on modeling and which we will also have the opportunity to discuss again. The different adaptations noted in the previous sections arise from different needs. It may have happened that a concept necessary for the study (instrumental conflict, for example) could not be studied because of its difficult accessibility (or its poor readability) through the model used. On other occasions, it has been difficult for us to represent communities of actors that are very different from each other (in a school DWS, for example, child, parents, teachers) acting together at the heart of a VLE. Moreover, the model on which we based ourselves had to be compatible with different points of view stemming from different theories developed around the instrumented activity (instrumental genesis, socioconstructivism, etc.): this constraint is another source of model evolution. Concerning Engeström’s model of analysis, all these transformations have concretely led to the development of the nodes of the model into clusters made out of subsystems, sometimes even by cloning the basic model several times, parallel to itself, as many divisions as there are communities. Educational innovations, including the arrival of active pedagogies in e-learning, have probably been the main cause of the many inflections of models. The latter, initially based on the provision of resources, were then replaced by models of pedagogical activities and finally encompassed all activities such as management, for example. Thus, we have moved quite quickly from a learning scriptwriting model to a training device model integrating all the activities that prevail there. It is obvious that we could also have evoked here more contemporary contextual changes that are, for example, at the origin of adaptive models, whose learning paths change dynamically, in response to the expectations of each learner [ZNI 12, BLA 13]. However, the current work targets, apart from in a few minor details, the same objectives as those targeted by the aforementioned researcher teams, and for that reason we prefer not to explore the subject further. We could also have talked about the work on the modeling of MOOCs that aims to support a personalized learning like those of Henning et al. [HEN 14] or those in progress led by the Canadian team [BEJ 15] which we recently heard a report about during a STEF laboratory seminar at ENS Cachan in June 2016. We will come back to this when we describe in more detail the context in which the MOOCs have established

134

Modeling of Next Generation Digital Learning Environments

themselves in France and worldwide. It is worth noting that these recent works “are based on the principles of open pedagogy and self-management of learning and pedagogical engineering methods (in particular, the engineering method of MISA learning systems)” [BEJ 15]. This work aims in particular at personalizing learning in a MOOC. We will return a little later to some of this work, but the aim in this section was above all to show that a change in socio-technical context, such as the arrival of a technopedagogical innovation, for example, can exert pressure capable of profoundly modifying the structure of the model in place. Before introducing VLEs into the paradigm of systemic modeling of complexity, let us consider that the goal mentioned previously is achieved.

4 The Digital Learning Environment in the Paradigm of Systemic Complexity Modeling

We have just highlighted the fragility of the current models when external conditions change, especially when these models are supposed to represent environments having undergone significant technopedagogical transformations. Before advocating the application of the complex systems theory to the modeling of new-generation DLEs (section 4.2), we will first clarify the founding concepts of this theory (section 4.1). It will be easier to later make a reference to argue this choice. 4.1. Complex systems theory and theoretical framework According to the theory of relativity and quantum physics, the complex systems (and chaos) theory is probably the third scientific revolution of the 20th Century. It drastically changed our way of thinking, perceiving the world and predicting the effects of such an event on the evolution of a system, in other words, how we do science. It is also referred to as systems science or complexity science. These new sciences “take their scientific status within the paradigm of constructivist epistemology; while normal sciences usually more readily refer to the paradigm of positivist epistemology” [LEM 99, p. 22]. There are many disciplines that one way or another refer to the constructivist paradigm. Among them and beyond, system and complexity science include education and communication sciences. For the latter, “the paradigms set up can be common to social sciences, either specific to communication sciences, such the scientificity of the approach, the positioning of man at the center of reflection, the Modeling of Next Generation Digital Learning Environments: Complex Systems Theory, First Edition. Marc Trestini. © ISTE Ltd 2018. Published by ISTE Ltd and John Wiley & Sons, Inc.

136

Modeling of Next Generation Digital Learning Environments

systemic nature of the objects analyzed or the constructivist dimension of reality, to name a few” [VIA 13, p. 4]. For scientific education, the constructivist paradigm can also be the basis for different theories of learning or knowledge development, and various educational intervention models may result. “It especially stems from Piaget’s epistemological and psychological dual perspective and Bachelard’s historical-criticism perspective from which didactics have borrowed the epistemological obstacle concept. These epistemological perspectives take the opposite view of positivism by emphasizing the role of the subject in the development of knowledge and on the impossibility to separate the facts from the interpretation models. Various learning theories can be registered in a constructivist paradigm while offering different explanatory models to account for the processes involved in learning: mental processing of information from the perspective of cognitive psychology; process of assimilation, accommodation and equilibration in the Piagetian perspective; process of gradual internalization of regulatory mechanisms of action and thought in the sociocultural context of Vygotsky, etc.” [LEG 04, p. 70] The common characteristic between these different “constructivist visions” is based on the idea that knowledge of the phenomena is the result of a structure made by the subject. The image of reality that the latter has or the notions that build this image are “a product of the human mind in interaction with this reality, and not a true reflection of reality itself”1. With this theoretical background having been specified, let us return to systems and complexity science before addressing more specifically the paradigm of systemic modeling of complexity. For Aurore Cambien, author of an impressive Certu study report2 on how to comprehend complexity: “The birth of systemic thinking is intrinsically linked to the emergence of a reflection regarding the notion of complexity during the 20th century. Awareness of complexity in the world goes hand in hand with that of so-called classic thought deficiencies to offer the

1 “Constructivist epistemology”, Wikipedia (2018, updated 20 March), available at: https://en.wikipedia.org/wiki/Constructivist_epistemology). 2 French center for studies on networks, transport, urban planning and public buildings.

The Digital Learning Environment in the Paradigm of Systemic Complexity Modeling

137

necessary means for comprehension and understanding of complexity on one hand, and action in response complexity on the other.” [CAM 07, p. 9] 4.1.1. Definition of a complex system A complex system is first and foremost a system. In fact, it consists of a set of elements interacting with each other and with the environment in order to produce the corresponding services to achieve its purpose. Beyond this overall systemic definition, complex systems are also, in essence, “open and heterogeneous systems in which interactions are linear and whose overall behavior can be obtained by a simple composition of individual behaviors” [ANI 06, p. 3]. Most of the formal properties that characterize a system (totality, feedback, equifinality, etc.) have already been defined and studied in section 1.2 in Chapter 1, when we discussed the interest in considering a DLE as a system of instrumented activity. This gave us the opportunity to return to the “General Systems Theory” formalized after World War II by Von Bertalanffy [VON 56]. Although these properties are still valid and active in complex systems, we will obviously reconsider their definition. In this section, instead, we will exclusively discuss the system’s complex nature. The word complex comes from the Latin word complexus meaning “what is woven together” and complecti, “something containing different elements” [FOR 05, p. 16]. Humans are rarely able to spontaneously understand complexity and it must be modeled and simulated in an attempt to identify it. Edgar Morin said that complexity exists when our level of knowledge does not allow us to tame all the information. “A complex system is, by definition, a system built by the observer who is interested in it. This said person postulates the complexity of the phenomenon without forcing himself to believe in the natural existence of such a property in nature or reality. Complexity, which always involves some form of unpredictability, cannot easily be held for deterministic purposes. In general, however, complexity is represented by a tangle of interrelationships interacting.” [LEM 99, p. 24]. In his book, La Méthode, Edgar Morin [MOR 08] helps us to understand complexity: he reveals to us the underlying concepts, requirements, issues and challenges. “The eminent thinker displays knowledge of complexity to command a

138

Modeling of Next Generation Digital Learning Environments

new approach and thinking reform” [FOR 05, back cover]. In fact, we can expect, following the work of Atlan [ATL 79] and Morin [MOR 08], complementing those of Varela [VAR 79] and Von Foerster [VON 60], that scientific advances in the field of design and analysis of complex systems are better taken into account. Aware that a universally accepted definition of the complex nature of a system has yet to be formalized, we will study it based on four features that create general consensus3: – the nature of its constituents (their type, their internal structure); – its open and chaotic appearance (outside influences on different scales); – self-organizing and homeostatic (features) dominated by interactions and feedback (non-linear interactions, often of different types); – its ability to shift to new systems with emerging and unforeseen properties [MIT 09]. Finally, we will address the epistemology to which these systems adhere to before differentiating the algorithmic complexity of natural complexity. 4.1.1.1. Variety of components and links The constituents of a complex system are of varied size, heterogeneous in nature (individuals, artifacts, symbols, etc.) and indistinct (possible evolution over time) and have an internal structure. “A complex system does not behave as a simple aggregate of independent elements, it constitutes a complete coherent and indivisible structure” [TAR 86, p. 33]. This formal property brings us back to the entire concept that we discussed in section 1.2.2 in Chapter 1. “The two notions that characterize complexity ‘are the variety of elements and interaction between these elements’. A complex system is formed by a wide variety of components organized into arborescent levels. The different levels and components are interconnected by a variety of links.” [TAR 86, p. 33] And it is the variety of interactions between its constituents and the variety of the constituents themselves that contribute the overall evolution of complex systems.

3 Opening speech of the International Conference on Complex Systems (International Conference on Complex Systems ICCS’12), organized in tribute to the late Professor Lorenzo Ferrer Figueras of the University of Valencia, Spain, available at: http://AES.UES-EUs.EU/.

The Digital Learning Environment in the Paradigm of Systemic Complexity Modeling

139

4.1.1.2. A complex system is not a complicated system A complex system is unpredictable. Therefore, it does not allow (by calculation), as forceful as it may be, a predicted outcome of the processes or phenomena4. But complex doesn’t mean complicated. “For example, we are able to understand and predict when and why a car engine breaks down while we are unable to understand and predict all of the factors that influence a change or a social movement (e.g. for a revolution).” [BEG 13] When the system is complicated, we are able to find one or more solutions, because a “complicated” system consists of a set of simple things. It just takes a bit of precision, knowledge and time to understand its function. The general method is first to break down and isolate each of the elements (in simple components or in simple notions) to then be able to reassemble them. “The mechanical function of a clock or a car engine, for example, may seem complicated to us but we are able to clearly explain how it works, because we know what impact each mechanical or electronic component has on the other” [BEG 13]. To illustrate the difference between a complex system and a complicated system, Snowden [SNO 02, p. 13] pertinently noted that “when rumors of reorganization appear in an organization, the human complex system begins to mutate in unpredictable ways and forms trends in anticipation of the change. Conversely, when one approaches a plane with a toolbox in hand, nothing changes!”5 4.1.1.3. Open system: between complexity and chaos A complex system is also an “open system” or in constant contact with its environment. In fact, “it undergoes external disturbances that are preconceived as unpredictable and unanalyzable. These disturbances, occurring in the environment, cause system adaptations that bring it back to a stationary state” [SNO 02, p. 32]. For those who study it, a closed system is more predictable than an open system in the sense where being totally isolated from external influences, it will be subject to internal modifications and will be found in a number of states defined by its initial conditions. And if the latter are invariable regardless of the number of experiences, in other words, the number of times that you observe the system’s behavior, from its initial 4 If this prediction is impossible by calculation, it becomes probabilistic with other tools such as systemic modeling of complexity. We will see this in the following section. 5 Author’s translation.

140

Modeling of Next Generation Digital Learning Environments

phase to its final phase (which is very difficult to attain), it will behave as an automaton that develops a known program. In this sense, it can be considered to be complex, as complicated as this program can also be! The knowledge of the various intermediate states being sufficient to define the system, its evolution and its finality become foreseeable and predictable. But as we will see below, it is difficult to ensure the invariance of the initial conditions applied to a closed system, which makes an open system and therefore a potentially complex system (because openness is a necessary condition but not sufficient to ensure that a system is complex). 4.1.1.3.1. Sensitivity to initial conditions In his time, Edward Lorenz raised the difficulty of such predictions when he sought to predict long-term weather events. Thanks to computer systems, and his knowledge of the deterministic laws of Galileo and Newton (also known as forecasters), according to initial conditions that would allow him to determine the future state of a system, he was given the objective of predicting the weather using computer simulations based on a mathematical model. In 1963, his work revealed the fact that tiny variations between two initial situations could lead to final situations that are unrelated. He claimed that it was not possible to correctly predict the weather over the long term (for example, one year) because an uncertainty of a millionth when entering data of the initial situation could lead to a totally incorrect prediction. The “butterfly effect” is a famous metaphor that illustrates this observation. Noted as part of “chaos theory”, it is intended to describe the fundamental phenomenon of sensitivity to initial conditions. This metaphor is sometimes simply expressed using a question: “Does the flap of a butterfly’s wings in Brazil set off a tornado in Texas?” This sensitivity can be translated by the following observation: no matter what the level of accuracy of the initial conditions, the system will nevertheless evolve in completely unpredictable ways. 4.1.1.3.2. Complex or chaotic system? The sensitivity to initial conditions is not the only responsible cause of the difficulty in predicting a complex system’s behavior. In an open system, the reciprocal influences between the activity of the system, its environment and its purposes continually alter its structure. “Random sounds” from the environment or the constituents of the system are active to the relationships that bind them. These events which are difficult to control (randomness of sound) are disturbances that contribute to changing the system. Whether they are internal or external, sounds

The Digital Learning Environment in the Paradigm of Systemic Complexity Modeling

141

alter the latter by evolving new forms and most are often unknown. Whether these systems are social or material, natural or artificial, it is very difficult to predict their behavior and their evolution because the world in which we live is mainly chaotic in nature and very rarely allows us to establish the system equations we want to study. Another reason why it is difficult to predict the behavior of a complex system is the ignorance of the rules that tell us how the system will change, in other words, the impossibility of modeling this system using predictive, solvent equations. But then, what good is there in studying a complex system if you cannot, in any reasonable way, predict its future behavior? The answer lies in the title of Henri Atlan’s book [ATL 79]: Entre le cristal et la fumée [Between the crystal and the smoke]. By this metaphor, the author tries to situate the complexity between these two extremes: the most certainly known crystal model and that of unpredictable smoke [ATL 79, p. 5]. 4.1.1.3.3. Algorithmic complexity and natural complexity We can differentiate algorithmic complexity from natural complexity. The first is one that we can process with a computer, a program or an algorithm. The goal to be achieved is known, as complex as it may be. The purpose is specified, and it determines the value of the procedure. Natural complexity is inherent to “systems not totally controlled by man because they are not built by man” [ATL 79, p. 5]: biological systems such as memory and social systems. This distinction is fundamental to the extent where the theory of self-organization that we will discuss is only justified in systems of natural complexity. These are a mix of knowing and ignorance. They constitute a sort of “compromise between two extremes: a perfectly symmetrical repetitive order whose crystals are more classic physical models and an infinitely complex and unpredictable variety in details such as evanescent forms of smoke” [ATL 79, p. 5]. Simon, 1974, has already shown the difference of intelligibility between natural and artificial objects in 1969. Natural objects are analyzed with the assumption that “complexity is woven from simplicity; correctly analyzed complexity is only a mask concealing simplicity”. Therefore, it would be appropriate, according to Simon, to find the ordered shape that hides in the apparent disorder. The theory of fractals, introduced in 1975 by Benoît Mandelbrot, would be a recent illustration. “The Santa Fe Institute6 was created in 1984 in an attempt to study the conditions by which order occasionally emerges from chaos, which is exactly the opposite of a butterfly effect”7.

6 http://www.santafe.edu/. 7 “Butterfly effect”, Wikipedia (2018, https://en.wikipedia.org/wiki/Butterfly_effect.

updated

31

August),

available

at:

142

Modeling of Next Generation Digital Learning Environments

Let us go back and take a quick look at the foundations of information theory [SHA 48] to examine this issue in more detail. According to this theory, all communication processes can be represented by a channel of communication between a transmitter and a receiver in which a message pass. The latter may be subject to external disturbances that generate errors, which modify the message. We call “sound” a disruption that modifies a message between its transmission and reception. For example, the letter or the symbol of an alphabet composing a loud message is received with a certain probability, but not certainty. This sound in the channel of communication has a negative effect in the formalization of the amount of information developed by Shannon [SHA 48]. However, instead of considering the transmission of information between the sender and the receiver, Atlan [ATL 79, p. 46] invites us to observe the amount of total information in the system (S) including this transmission path. In this case, the relevant transmission path is no longer the same. It settles between the system and outside observers.

Figure 4.1. Zoom on a color representation of the Mandelbrot set. For a color version of this figure, see www.iste.co.uk/trestini/learning.zip

The Digital Learning Environment in the Paradigm of Systemic Complexity Modeling

143

Figure 4.2. Message transmitted in a channel that leaves the system and leads to the observer

From this perspective, the ambiguity introduced by sound on a basic channel (the one seen by Shannon) can be a generator of information for the observer. Intuitively, we perceive that the complexity notion of a system is intimately linked to the amount of information defined by Shannon. “The more a system is composed of a large number of different items, the greater its quantity of information and the improbability of it forming ‘as it is’ by randomly assembling its constituents. That is why this quantity could be proposed as a measurement of the complexity of a system, in that it is a measurement of the degree of variety of elements that constitute it” [ATL 79, p. 45]. Complexity is often associated with newness (high amount of information), while the trivial is repeated (small amount of information). “The quantity of information of a system, made up of parts, is then defined from the probabilities that we can assign to each kind of component on a set of supposed statistically homogeneous systems from each other; or even from all the combinations, where it is possible to achieve with its components, that form possible states of the system.” [ATL 79, p. 45]. Errors caused by sound introduce a new variability within the system. Thus, the increase in the degree of variety (or degree of differentiation or complexity) of a system, based on a random disturbance from its environment, is measured by the increase in the amount of Shannonian information that it contains.

144

Modeling of Next Generation Digital Learning Environments

Henri Atlan transposes this increase in the amount of Shannonian information in a loud system as far as the complexity of an organized system, such as biological organisms or a social organization. This conversion leads the author to design “the evolution of organized systems or the phenomenon of self-organization as a process of increasing both structural and functional complexity resulting from a succession of disruptions caught, each time followed by a recovery to a level of greater variety and lower redundancy” [ATL 79, p. 49]. The increase in complexity “can be used to achieve greater performances, most notably with regard to the possibilities of adaptation to new situations, thanks to a variety of responses to diverse stimuli and randomness in the environment” [ATL 79, p. 45]. In fact, a very organized system, such as an algorithmic complexity system, for example, will be totally predictable and, consequently, poor as to its content information in a Shannonian sense. Otherwise, if it is completely erratic and random, its predictability will be void. It is therefore not complex, but chaotic. “It is when the content information is intermediate in terms of predictability and average probability, that complex systems can be found.” [WAL 12, p. 1023] It is in this limited space that it is possible to predict potential emerging phenomena with, undoubtedly, a certain degree of uncertainty. Other definitions of complex systems or associated concepts, mainly from thermodynamics, also exist and can be used depending on the circumstances, as we will see below. 4.1.1.4. Feedback, self-organization and homeostasis 4.1.1.4.1. Feedback A fundamental feature of a complex system (already studied above) is its circular causation. In more explicit terms, individual or collective constituents of the system behavior provide feedback on their own behavior. These elements will individually or collectively change their environment, which in turn will constrain them and change their states or possible behaviors.

Figure 4.3. Feedback process of the effect to the cause

The Digital Learning Environment in the Paradigm of Systemic Complexity Modeling

145

This circular process applied to systems, as opposed to the linear and unidirectional Shannon model, is in the foundations of social theory of communication. Expanding cybernetic thinking to all living systems, which our learners are a part of, Bateson [BAT 77] uses this concept of feedback to explain a very broad set of phenomena, such as language, biological evolution and schizophrenia… but also learning. Thus, it deals with learning as a feedback phenomenon: “change can define learning”, he says. For him, “the word ‘learning’ undoubtedly indicates a change of one kind or another” [BAT 77, p. 133]. What is essential in a complex system is therefore not so much the number of factors or dimensions (factors, variables) related to the system constituents, but the fact that “each of them indirectly influence others, who themselves influence it in return, making the system’s behavior an irreducible whole”8. In other words, in a complex system, knowing the properties and behavior of isolated elements is useful but not sufficient in order to predict the overall behavior of the system. To predict this behavior, it is necessary to take them all into account, which usually involves making a simulation of the studied system, a simulation we will look at in detail in section 6.2 in Chapter 6. 4.1.1.4.2. Self-organization We have just explained that a system can be subject to external disturbances that generate errors, which modify it. These disturbances are called “sounds”. When the system is reached by a sound from the environment, it can be the source of an increase in the complexity of the system. Initially, the system is disrupted and then it recovers by increasing its level of variety. As a system, it can expand this disorganization. The increase in the complexity of a system can produce a positive effect on it. The increase of complexity “can be used to achieve greater performances, most notably with regard to the possibilities of adapting to new situations, thanks to a variety of responses to diverse stimuli and randomness in the environment” [ATL 79, p. 49] Accordingly, when a random sound from the environment reaches the system, it integrates to rebuild in a new way. This is true regardless of the system, on the condition that it is natural complexity (cognitive system, memory, DLEs, etc.). If this mechanism is unknown to us in detail, it allows the system to do its own learning by looking to increase its degree of complexity (or variety). It is also the reason why this sound is said to be organizational. 8 “Complex System”, Wikipedia (2018, updated 8 June), available at: https://en.wikipedia.org/wiki/Complex_system.

146

Modeling of Next Generation Digital Learning Environments

4.1.1.4.3. Homeostasis In section 1.2.3 in Chapter 1, we defined homeostasis as a resistance to change specific to all complex systems. This can now be explained using self-organizing capabilities. Social systems such as ecological and biological systems, and even more so education, are particularly homeostatic systems. “They oppose change by all means at their disposal. If the system fails to restore its balance, it then enters a mode of operation with, at times, more drastic constraints than the previous, which may lead, if the disturbances continue, to the destruction of the whole group.” [DER 75, p. 129] Atlan [ATL 79, p. 47] formally shows by what mechanisms these critical cases can occur. After a fairly long period of development, he concluded that we are dealing with two kinds of effects of ambiguity generated by sound on the general organization of a system. He calls these effects of ambiguity “destructive ambiguity” and “ambiguity-autonomy”, the first being considered negatively, the second positively. A necessary condition so that the two coexist would be what Von Neumann [VON 66] called an “extremely highly complicated system” and that Edgar Morin [MOR 73] called subsequently, probably more correctly, a “hyper-complex system” [ATL 79, p. 48]. Between these two extremes, “homeostasis appears as an essential condition for stability and therefore survival for complex systems. Homeostatic systems are ultra-stable: all of their internal, structural, functional organization contribute to the maintenance of this same organization” [DER 75, p. 129]. 4.1.1.5. Existence of emerging properties Perpetual reorganizations of systems can sometimes produce unexpected and unpredictable behaviors. We say that such behavior is “‘counter intuitive’ in the words of Jay Forrester or counter-variant: when following a specific action, we expect a given reaction, although a completely unexpected and often an opposite outcome is often obtained. These are the games of interdependence and homeostasis. Politicians, businessmen or sociologists don’t know much about the effects” [DER 75, p. 129]. If there is one thing that characterizes many complex systems, it is certainly the existence of emerging properties, either in the living or inert world. These phenomena, which are also emerging, are well known in the world of matter. Atoms, for example, bind to produce a new entity with new properties:

The Digital Learning Environment in the Paradigm of Systemic Complexity Modeling

147

the molecule. For example, who could have predicted that from gas molecules, such as dihydrogen (H2) and oxygen (O2), we could produce a liquid substance with completely new characteristics, water (H2O)? In addition, these water molecules crystallize in the cold to produce an unpredictable new compound, ice crystals… The first conclusion that can be drawn from these phenomena is that it is difficult to predict the overall behavior of the system from the mere knowledge of the properties and behavior of isolated elements. A complex system has, in fact, new and higher properties than those that we would have obtained as the sum of the properties of its constituents. Let us also keep in mind from these examples that complex systems are not linear “since the new function or the new state does not appear gradually, but suddenly: water freezes, life appears” [WAL 12, p. 1022]. Many other emerging phenomena can be cited beyond the phenomena derived from the world of matter. For example, when we think about the human mind or conscience… Do they not emerge from a “simple” organic aggregate? “The obvious emerging phenomena that have been shown during the evolution of the physical and social universe are undeniably the appearance of life, the emergence of thought and the genesis of institutions (language, currency). However, we can observe a variety of emerging phenomena at a more modest pace such as turbulence in fluids, cell clusters, colonies of insects or the prices of goods. […] These emerging phenomena can take various forms, ranging from simple combinations of properties of microscopic entities or relations between these entities concerning the appearance of characteristics or original macroscopic entities.” [WAL 09, p. 2]. Thus, emerging phenomena can occur “through all natural sciences, such as physics, chemistry, biology, but also at the level of those dealing with the environment, climate, society, energy distribution systems and the internet” [WAL 12, p. 1022]. It then becomes possible to imagine that social interactions within digital learning environments from the last generation could also be the cause of emerging phenomena… Could we not imagine, for example, MOOCs quickly becoming real parallel universities that compete with conventional universities, calling into question the methods and teachings of the latter, even while MOOCs were designed to be at the service of universities? Alternatively, could we not conceive of the opposite assumption, whereby MOOCs would finally be abandoned by users, like a fashion phenomenon, even though they currently seem fated to success?

148

Modeling of Next Generation Digital Learning Environments

4.1.2. Modeling of complex systems “As soon as we exceed the description and classification for the analysis ‘of processes’, this implies a reference to a model, of which the definition of which should be more precise, as the reasoning that we apply to it is more elaborate.” [GOG 16] We discussed in section 2.2.1 in Chapter 2 the different functions of modeling as well as the nature of these models and their uses. In doing so, we were very quickly positioned in this field by vacating our immediate research concerns, models known as “theoretical” in order to only keep “object”-type models [LEG 93] and especially graphic and visual models. However, the latter allow the development of theories that bring back the phenomenon studied to a more general phenomenon, in agreement with experience and its confrontation [GOG 16]. Therefore, throughout our modeling approach, we will make reference to theories such as the modeling of complex systems, for example, [LEM 99] that we consider central (inspired by the previous theories, synthetic and relatively current). This is the reason why we refer to it frequently. However, from a pragmatic point of view, our models have no other ambition than to provide “a schematic representation of an object or a process that allows you to substitute a simpler system for a natural system” [GOG 16]. The theories solely help us to construct these models. In other words, our intention is not to construct a theory on the modeling of complexity, but to use the complexity theories to construct models that are expected to facilitate our analysis. EXAMPLE.– The simplification of the “object” model to which we have just referred can be done by performing a series of reductions and simplifications (well-thought-out) of information9 that the conjunctive logic of “modeling of complex systems” (theoretical model) encourages us to do. We will revisit this fundamental process of reduction and simplification of information (most notably through statistical methods) in great detail, that allows SM (systemic modeling, as opposed to analytical modeling, AM) in the section dealing with the arguments that lead us to apply the complex systems theory to the modeling of new-generation DLEs (section 4.2). Moreover, advancing through these different theories, we gradually found that a connection between the systems theory and the theory of modeling became essential to the understanding of complexity. Le Moigne [LEM 77, LEM 94] explains why in his book entitled, Théorie générale des systèmes [General theory of systems]:

9 By factorial methods, for example.

The Digital Learning Environment in the Paradigm of Systemic Complexity Modeling

149

“The ‘General System’ concept was precisely forged to facilitate the apprehension, or understandings, of open processes rather than closed objects. Not a preconceived premise of existence (or rejection) of systems in nature (nature of things or human nature), but a project with lucid understanding: to represent the phenomenon considered as so by a system in general. This taking advantage of the accumulated modeler experience has taken place since the rhetoricians of Ancient Greece (‘inventio’ rhetoric) up to Condillac’s ‘Treaty of systems’10. By accepting this well thought-out and reasoned formulation of a ‘systems science’ project, we can legitimize and ensure this argument: ‘Systems science is understood as the science of systemic modeling’.” [LEM 94, p. 7]. DEFINITION.– A complex system is a model of a phenomenon perceived as complex that is built by systemic modeling [LEM 99, p. 41]. In a second book on the modeling of complex systems, this same author presents the classic methods of modeling and highlights their failures. Through examples, parables and historical reference points, he shows the limits of these models over time. “The question posed by the author is whether one can truly solve complex problems with analytical models. To address this issue, he begins by distinguishing complexity and complication. It starts from two postulates: – the simplification of applied complications to the complex results to an increase in complexity; – the modeling system projects are not given: they are built. Next, modeling is defined. He introduces the notion of intelligibility that does not have the same significance for a complicated system (explanation) in a complex system (understanding) […]. In other words, the author shows how, in the face of a problem, we develop models with which we reason.” [NON 00, paragraph 2]

10 See the article “Système” from the Dictionnaire d’histoire et de philosophie des sciences, 1999, [PUF 99, p. 901].

150

Modeling of Next Generation Digital Learning Environments

The systemic modeling of complexity is developed throughout the book, referencing the contributions of several thinkers such as Jean Piaget11 and Paul Valéry Goethe, who are not satisfied with analytical models. The limits of analytical thinking facing complexity are also highlighted in numerous works and in varied disciplines such as biology, mathematics and neurophysiology, considered as being the origin of the development of systemic thinking [CAM 07, p. 9]. 4.1.2.1. Conjunctive versus disjunctive logic “The analytical method needs disjunctive logic since the results of the breakdown must be definitively distinguished and separated” [LEM 99, p. 32]. However, according to the author and all the thinkers he cites, this logic does not, “account for phenomena that we perceive in and by their complex conjunctions…” [LEM 99, p. 33]. He blames in addition the analytical school supporters “seeking to adapt the problem available to models rather than seeking models of substitution that best meet the problem” [NON 00, paragraph 2]. The use of analytical models requires a simplification of the studied phenomenon, so when a breakdown or division sets off from the hypothesis, in order to address a complex situation, the register must be changed. It is required to go from the registry of disciplined knowledge review to that of active enlightening knowledge methods. To guide and argue his reasoning, Le Moigne refers to some stable benchmarks which he called axioms. He thus said gather “the contemporary state of maturation of axiomatics based on the texts of Heraclitus, Archimedes, Leonardo da Vinci, GB. Vico or Hegel” [LEM 99, p. 35], in the form of three axioms that fit well within a constructivist epistemology (the paradigm of the constructed universe). In this argument, this radically comes in opposition with positivist epistemology (the wired universe paradigm) that always relies, according to the author, on Aristotelian axioms of analytical modeling and disjunctive logic. Let us quote the three axioms of systemic modeling referred to by the author [LEM 99, p. 36]: – the axiom of teleological operationality or synchronicity: a modelable phenomenon is perceived as an intelligible action and therefore teleological (not erratic, showing some forms of regularity); – the axiom of teleological irreversibility (or diachrony): a modelable phenomenon is perceived, transforming and project forming over time;

11 “In the 1980s, Jean Piaget proposed designing a ‘logic of significations’, something being developed at the Geneva School (R. Garcia, 1987); but the discussion of its axiomatic foundations remained to be shown” [LEM 99, p. 32].

The Digital Learning Environment in the Paradigm of Systemic Complexity Modeling

151

– the axiom of inseparability or recursion (or excluded outsiders, or conjunction, or autonomy): a modelable phenomenon is perceived inseparably joining the operation and its product, which can be a producer for itself. Referring to these three axioms, it appears that we can actually ensure the precision and cognitive consistency of modeling reasonings [LEM 99, p. 35]. Each new theoretical contribution is then associated with “a canonical model that represents the new system” [NON 00, paragraph 2]. The first is the canonical form of the general system that leads the author to define four founding concepts, grouped in pairs in procedures: the cybernetic procedure includes active environment and project concepts, while the structuralist procedure includes synchronic operation and diachronic transformation. The inseparability of these concepts leads to the conceptualization of the general system, understood as “the representation of an active phenomenon perceived identifiable ‘by’ its projects in ‘an’ active environment in which it functions ‘and’ transforms teleologically” [LEM 99, p. 40]. At the beginning of Chapter 5, we will revisit the concepts underlying cybernetic and structuralist procedures. In a more mnemonic way, “the general system is described by an action (a tangle of actions) in an environment (‘carpeted with process’) for some projects (purposes) functioning (doing) and transforming (becoming)” [LEM 99, p. 40]: therefore, for the author, the basic concept of systemic modeling is not its structure, but the action it represents by a black box or a symbolic processor that takes account of the action. “Modeling a complex system first involves modeling a system of actions.” [LEM 99, p. 45]. 4.1.2.2. A complex system and canonical model process The modeling of a complex action is characterized by a general process notion, which is defined by its exercise and result. The author represents the process using three functions: the temporal transfer function, the function of morphological transformation and spatial transfer (see Figure 5.21). “We define a process by its exercise and result (an ‘implex’ therefore): there is a process when there is, over time (T), modification of the position in a ‘Space-Form’ reference, and a collection of ‘products’ identifiable by their morphology, and thus by their form (F). The preconceived conjunction of a temporal S transfer (moving in space: transport, for example) and a temporal F transformation (morphology modification: an industrial treatment

152

Modeling of Next Generation Digital Learning Environments

transforming flour and water into bread, for example) is by definition a process. Its outcome can be recognized: a movement in a ‘Time-Space-Form’ reference, identified by its exercise.” [LEM 99, p. 47] In this regard, Martin et al. noted: “In this theory, the observed system is apprehended in the form of a collection of tangled actions, the process notion being conceptualized by that of action. If the process is a formal construction expressed in the language of mathematics (differential equation) in ‘Time–Space–Form’ references, the description of the action expresses itself in natural language in response to the question, ‘What does that do?’ Although this theory is based on the action concept, no formal description framework of an action is proposed (Larsen-Freeman and Cameron, 2008).” [MAR 12] We will see later that these authors propose a formalism for the action concept we want to deepen and that will be very useful for modeling DLEs considered as complex systems. 4.1.2.3. Complex system and action interrelation Returning to feedback concepts, Le Moigne characterizes each processor by its “inputs” and “outputs”, at every “ti” moment, before establishing three archetype processors that he adapts specifically into three transfer functions: temporal, spatial and morphological. He also distinguishes processors mainly processing symbols (or information) from others (mainly processing tangible goods) [LEM 99]. He then takes an interest in cases where the number of considered processors becomes high (frequent cases in complex systems) and then proposes differentiating the system into “many subsystems or ‘levels’, each level with the capability to be modeled by its network and relatively interpreted independently as soon as the inter-level coupling interrelationships have been carefully identified” [LEM 99, p. 53]. “We’ll have the opportunity to show [he says] that this deliberate a priori articulation of a complex system ‘in multiple functional levels’ (by intermediary projects) will lead to a very general canonical model which will often lead in turn to an intelligible modeling of its organization.” [LEM 99, p. 57]

The Digital Learning Environment in the Paradigm of Systemic Complexity Modeling

153

4.1.2.4. An archetype model of the articulation of a nine-level complex system K.E. Boulding [BOU 04a] and subsequently J.-L. Le Moigne [LEM 99, p. 58] offered a definition of an organization based on the definition of increasing levels of complexity of a system, each level also containing all the characteristics of lower levels. The article ‘General System Theory, the Skeleton of Science’ (the “framework of science”, [BOU 04a]) presents Boulding’s famous nine levels [BOU 04a] analysis, and we will quickly present it here. Keep in mind the significance of each of these levels when applying the approach to DLE modeling that we saw in Chapter 5. We will avoid redundancy by this time citing the way Le Moigne has interpreted them. What makes this nine levels analysis important is not so much the manner in which it is formulated, but the way it has contributed to “clarifying the terms of use of the systemic approach or methodology”12. We will only cite briefly the first six levels, referring the interested reader to the description of other levels, either by the original document (op. cit.), the work of Tardieu, Nanci and Pascot [TAR 79], or the reinterpretations made by Le Moigne [LEM 99] in Chapter 5. – Level 1: “The first level is that of static structure”13 [BOU 04a, p. 202]. The phenomenon to model is identifiable but passive and without need. – Level 2: The phenomenon is active. It does, meaning it “‘transforms’ one or more ‘inflows’ into one or several ‘outflows’” [TAR 86, p. 60]. – Level 3: It is active and regulated. “It presents ‘patterns of behavior’”. – Level 4: It inquires. “The information allows a processor to know something about the activity of another processor in the system or its environment. The regulation of a system, thanks to a flow of information, is the basic concept of cybernetics” [TAR 86, p. 60]. The system transmits regulation instructions, reflecting changes in the system’s state. – Level 5: It determines its activity. This level of complexity suggests links that will be established between a “control system” and an “operating system” that we will later define. Here, the system is seen capable of making decisions concerning its own activity.

12 “Kenneth E. Boulding”, Wikipedia (2018, updated 29th July), available at: https://en.wikipedia.org/wiki/Kenneth_E._Boulding. 13 “The first level is that of static structure” [BOU 04a, p. 202].

154

Modeling of Next Generation Digital Learning Environments

We will come back fairly quickly to this fifth level of organization and cover level 6 in the section relative to the concept of organization as well (section 4.1.2.5). Note that at this time the definition of tree diagram concepts (seen in section 4.1.2.3) and the concepts of organization levels proposed by Boulding (seen in section 4.1.2.4) is a critical first step in modeling complex systems. “From the first level corresponding to static and simple objects of physics and chemistry up to the last level of socio-culture, the movement is that of increasing complexification. The understanding of the system represented by the ninth level involves all the previous levels.” [CAM 07, p. 16]. “This presentation of complexity, in addition to its educational interest, offers a methodological interest. It illustrates the specific approach of a speaker (or curious person) when becoming acquainted with the object information system […]. Social organizations, the goal of our modeling, are at the highest level; the ninth.” [TAR 86, p. 59] 4.1.2.5. Active organization: operating system and control system If there is a central concept that allows us to carefully describe a complex system, it is that of active organization. “Say that a business, a municipality, a family, a government, an industry, an ecosystem, a university, a hospital, a club, etc., are complex systems. This is already the application of some strong assumptions that we’ve recognized. It relates to a complexity of irreversible, recursive and teleological actions that we propose designating (or modeling). The organization of this modeling exercise is through a pivot concept (today very well clarified) known as active organization (that E. Morin proposes calling Organis-action and F. Perroux the Active Unit).” [LEM 99, p. 73] The organization is here considered as “active” because it is the one who determines its own activity. If the behavior of the system was solely dependent on external interventions, it would not be complex but simply predictable. The complexity appears with the emergence of autonomic capabilities within the system. The behavior of a complex system regulates endogenously; therefore, it is the one ensuring that its behavior is consistent with what was expected and eventually determines to act for modification. The organization is therefore made up of, “on the one hand, an ‘operating system’ (OS), the place of activity allowing transformation of incoming and outgoing flows and, on the other hand, a ‘control

The Digital Learning Environment in the Paradigm of Systemic Complexity Modeling

155

system’ (CS) capable of, in view of the information representing the activity of the operating system, reacting according to a certain project (certain objectives) to make decisions […]. The dynamics of the organization [active organization] concern, on one hand, ‘the functioning of the operating system’ and on the other hand, ‘the functioning of the management system’” [TAR 86, p. 61], as suggested in Figure 4.4.

Figure 4.4. Level 5 description of an organization: 14 the object determines its own activity

4.1.2.5.1. The functioning of the operating system Metaphorically, the functioning of the operating system can be represented by a succession of images of each input–output pair, taken “intermittently”, which represent the state of the system, at every successive moment, seen from the exterior. In fact, its function can be defined as “‘the chronicle of instantaneous observation’ by which the representation system15 has agreed to identify the object’s behavior” [TAR 86, p. 62]. However, this series of states accurately reflects the observed system function and not necessarily one that was expected. It then becomes useful to compare the observed states, at each instant “T” , to the desired operation. This leads us to consider, on the one hand, a functional structure resulting from a design (a program) that will model the desired operation, and on the other, a

14 Diagram created by us but strongly inspired by Figure 5 of Tardieu’s book, et al. [TAR 86, p. 62]. 15 We call a “representation system” the system that selects and builds information circulating within the organization; there is often mention concerning its reference.

156

Modeling of Next Generation Digital Learning Environments

model that will describe the observed behavior. Doing so, immersed in its environment, the operating system that will without a doubt be the target of unforeseen external disturbances (organizational noises) will account for the control system of these variations (representation information). It will then take the necessary decisions to ensure that the organization optimally follows the objectives having been set (decision information). 4.1.2.5.2. The functioning of the control system The control system therefore receives input from the representation information regarding the functioning of the output product and operating system of decisions viewed to maintain or change its behavior. This regulation will be based on the organization project, whose details must obviously be known. We can also expect that this verification be done at the control system level by controlling good understanding of the organization project. Therefore, the control system itself has a functioning that can be modeled “as a chronicle of observation” [TAR 86, p. 63] of couples (inputs and outputs) of its own system and its state. “This means that the association of representation information provided to the control system and decision information taken by it is a representation of the functioning of its internal logic” [TAR 86, p. 63]. Looking ahead in our future modeling, we would rightfully have to imagine that there is a link between the decision-making process to limit an MOOC to a six-week duration and the observed resilience of registered users. 4.1.2.6. Information system “Information allows the organization to adapt its behavior at every moment by regulating, transforming and rebalancing in order to be in harmony with the environment. Therefore, information leads to a process of permanent adjustment of the organization by channels (the system adapts by accommodation) and codes (the system adapts by assimilation) of communication with respect to a project.” [NON 00, paragraph 2] At the sixth level of organization, modeling the information system (IS) is introduced. It ensures a coupling between the information considered during its storage and information considered at the time of its use. In fact, “in developing its decisions, the system does not consider instant information, such as, for example, a conventional thermostat, or a reflex arc; it also considers the information it has stored” [LEM 99, p. 61].

The Digital Learning Environment in the Paradigm of Systemic Complexity Modeling

157

Figure 4.5. Level 6 description of an organization16

At this level, the control system cannot only have representation information on the current state of the operating system (for example, who is exchanging information on a discussion thread, looking back at our activity example on MOOCs), but also on the past states of the operating system (number of registered users having been active on the discussion thread during the last training session, for example). The questions that will now arise to complete level 6 concern, on the one hand, the dynamics in the organization’s information system (where system information/operating system interactivity exists) and, on the other hand, the dynamics of the operating and control system within the information system. We refer the interested reader to the details of these three representations in “the Merise method” [TAR 86, p. 65] seen in section 3.5 in Chapter 3. Nevertheless, and essentially: “The organization’s information system (IS): – records (in symbolic form) representations of the functions of the operating system (behavior of the complex system); – stores them;

16 Diagram created by us but strongly inspired by Figure 6 of Tardieu’s book, et al. [TAR 86, p. 65].

158

Modeling of Next Generation Digital Learning Environments

– makes them available, generally by means of an interactive form, to the decision-making system. Which, after having developed its action decisions (orders), also records and stores them through the IS, transmitting them ‘for action’ to the operating system.” [LEM 99, p. 87] 4.1.2.7. The canonical model of a three-level organization system: OS, IS and DS It is possible to represent the organization with an established model: – decision-making system; – information system; – operating system. Figure 4.6 shows a possible diagram of this organization. Le Moigne adds that “the canonical model of system organization, connecting the three levels: operation, information, decision-making, is universal. You can always start an organization’s modeling of a complex system using this genotype model” [LEM 99, p. 87].

Figure 4.6. Canonical model of a system and its three subsystems (OS, IS and DS)

The Digital Learning Environment in the Paradigm of Systemic Complexity Modeling

159

4.2. Argument in favor of the application of the complex systems theory to the modeling of latest-generation DLEs 4.2.1. Introduction In Chapter 3, it was clearly shown that there are limits to some design or DLE analysis models according to the context in which they are involved in. They are subject to relatively important sociotechnical transformations linked, for example, to the emergence of technopedagogical innovations. As an example, let us first emphasize Engeström’s model in order to recall the first modifications that we plan to make, discuss them and develop arguments in favor of a possible new path, like, for example, that of a radical change paradigm in terms of modeling. With the emergence of social networks, MOOCs, informal learning networks and connectivist learning approaches, we realized that digital learning environment (DLE) analysis had become even more complex to conduct than ever before. Let us look back on this significant development and the events that have accompanied it. After having developed each of the poles of this model “into clusters”, duplicated and stretched into the three dimensions of space within its internal structure and added new relationships between its constituents (see sections 3.2 and 3.3 in Chapter 3), a new sense of incompleteness arises concerning this representation. It then seemed important to reconsider the systematic use of this triadic model and even its expansion, where we again found little adaptation to these new study objects. This sense of incompleteness brought about a counter flow to the teachings of systemic modeling (see section 4.1.2) through its contemporary explanatory efforts: to reason in joining rather than separating. The model structure17 constitutes various divisions (subject, object, tool, community, work division, etc.), although practical in certain simple analysis cases, and shows (in the art of reasoning) the adoption of disjunctive logic specific to analytical modeling (AM)18. This finds itself in conflict with the conjunctive logic advocated by systemic modeling (SM). Therefore, in justifying a potentially infinite breakdown of system constituents-objects, we had the feeling of gradually distancing ourselves from the canonical form of a “general system” (see Chapter 4). 17 Here it relates to its original structure (the one proposed by Engestrom) and especially to the one which we expand on. 18 Even if Engeström asserts its systemic nature (see section 4.2.2.4).

160

Modeling of Next Generation Digital Learning Environments

Engestrom’s model [ENG 87] could certainly take advantage of the time period where new contributions to systems science were established. It could also identify with systematism as many other so-called “systems analysis” models developed in the present period (and previous years). However, in light of the work on systemic modeling of complexity, this model relies on a disjunctive axiom since the results of breakdown into the model’s constituent divisions must be definitively distinguished and separated. This impression of infinite decomposition into subsets of finer constituents (and not of functions or specialized processes) was always present at the time when we wanted to apply it to instructional design models (see section 2.4.1 in Chapter 2). Moreover, the latter have shown other inadequacies, for example, their difficulty to be able to adapt to certain distance education teaching courses, most notably [MEI 94] collaborative or situational problem learning [HEN 01]. Despite the enormous progress that has been made in this area, we had to accept that current EMLs were still struggling to properly represent collaborative activities built around situational problems, with evidence provided by the following authors’ works: Pernin and Lejeune [PER 04b]; Ferraris et al. [FER 05]; Laforcade [LAF 05a]; Laforcade et al. [LAF 05b]. Furthermore, the adoption of a more “macro” approach, to us, appeared more adapted in the context of the current research aimed at taking into account more successfully all the present activities at the heart of the DLEs [OUB 07, VIL 07, etc.]. The so-called education engineering has this goal: also taking into account management activities, for example (see section 2.4.2 in Chapter 2). Finally, the current adaptive models (see section 2.4.3 in Chapter 2) focusing particularly on technopedagogical innovations like MOOC [BLA 13, BEJ 15] aim to anticipate learners’ needs and to provide an answer, and this is regardless of the complexity of the system. However, in this specific case, are we not facing another challenge, namely that of wanting to design a system capable of adapting to all situations encountered? In the words of Atlan [ATL 79, p. 5], “in this type of system, the goal to achieve is known, as complex as it may be”. The purpose is specified and determines the value of the procedure. Here, we typically recognize what Atlan called an algorithmic complexity system. Natural complexity is inherent to “systems not totally controlled by man because they are not made by man” [ATL 79, p. 5]. Wanting to substitute templates specific to the theory of complexity in models of the Canadian team, LICEF, [BEJ 15], based on an ontology modeling, there is nothing provocative: it is simply a way to “replace the research accuracy of materialistic ontology by a more uncertain phenomenology constructing approximated models, but that better correspond to entirely real phenomena, and which reductionism cannot account for” [ZIN 03].

The Digital Learning Environment in the Paradigm of Systemic Complexity Modeling

161

As the organization of a complex system is considered “active” (it is a concept of SM), it is the one (and not a program) that determines its own activity (see section 4.1.2). If the system behavior was solely dependent on external interventions, it would not be complex and even less unpredictable. Complexity appears with the emergence of autonomic capabilities within the system. “In human groups, for example, and more particularly in the case of the emergence of panic or a rumor spreading among [PRO 07] crowds19, self-organization is not the result of a predetermined intention. Agents or entities in interaction, without a previously defined common goal, will create a certain form of organization without knowledge and by imitation. Therefore, what characterizes self-organized systems is the emergence and maintenance of global order without there being a conductor” [PRO 08, p. 2], and we assume that this upholding of order cannot be preconceived. The theory of self-organized criticality proposed by Per Bak, Chao Tang and Kurt Wiesenfeld in 1987 [BAK 96] shows that it is impossible to predict and anticipate all behaviors of a complex system. Just as useful as the “adaptive system” design (in the sense of education engineering), we also believe in their ability (in the sense of the theory of ‘adaptive’ complex systems) to better react to disturbances, both externally and internally or to self-regulate. As advised by Davis [DEL 00], we should not fear complexity: “In wanting to simplify reality, do we not risk putting ourselves on the fringes of the wealth it offers?” In summary, the hypothesis that we offer in order to justify the adoption of the paradigm of systemic complexity modeling is mainly due to the recognition of complex phenomena, which act at the very heart of our object of study (latest-generation DLEs) and by alleged inadequacy of analytical modeling (AM) to solve the problem posed. However, other reasons reinforce this deliberate choice, for example: – the existence of many convergence points between connectivist learning design and the complex systems theory; – the existence of a conceptual analogy between the SM of the complexity and SM of the activity; – the existence of languages “better” adapted to open complex system modeling, open in addition to social sciences; – to follow a forward-looking approach allowing answers to the many societal questions on the future of these digital environments, etc. 19 On 14 August 2016, we witnessed the explosion of firecrackers to simulate a terrorist attack at Juan-les-Pins, in the Alpes-Maritimes.

162

Modeling of Next Generation Digital Learning Environments

Below, we will develop and explain these few peripheral arguments, after having done the same thing with our two central arguments. 4.2.2. DLE: a complex system Looking back at the different characteristics of a complex system from section 4.1, let us explain what leads us to consider a latest-generation DLE as a complex system. 4.2.2.1. A world of knowledge: a complex system To have the project of modeling a world of knowledge, such as latest-generation DLEs, referring to the constructivist paradigm of systemic complexity modeling, may seem like an audacious exercise. But moreover, assuming that all acquisition of knowledge can adapt to a world not necessarily built for the sole purpose of learning may prove to be the challenge. However, we wanted to address this challenge by perceiving cognitive organization and the human cognitive system as self-organizing complex systems capable of learning in a world organized for this purpose (rather than built in the formal sense of the term), by combining two shifts of thought of which the union seemed, only a few years ago, utopian. Let us argue the choice here, when the main development is (at the personal request of Jean-Louis Le Moigne) in the “Cahier des Lectures MCX (Modélisation de la Complexité)” of the “Réseau Intelligence de la Complexité” [TRE 09b]. The first of these shifts is based on the complex systems theory, which assumes that any learning opportunity results from the disruption of a system subject to external disturbances. Whether this system be social or cognitive, it then reorganizes by increasing its degree of variety or complexity. “[As a system, it is enriched by this disruption by the means of] a non-directed learning process, in the sense where this learning is by no means a program established in human memory, or in the natural or social environment of this memory. […] The efficient cause of this learning is the random encounter of system memory and noise factors from its environment. […] This learning product is a succession of psychological categories always finer or differentiated, whose list and construction methods are likely to be involved at any stage of the process.” [ANC 92] We can intuitively understand this phenomenon when thinking of the classical theory of evolution: “Mutations, which are precisely DNA replication errors, are considered to be the source of the progressive increase of the diversity and complexity of living human beings” [ATL 98]. The increase in complexity in a

The Digital Learning Environment in the Paradigm of Systemic Complexity Modeling

163

system can produce a positive effect on them; it becomes a learning source. However, this theory is also taken for granted in the sense that this self-organizing process can only operate in systems with natural complexity, or inherent complexity, to “systems not totally controlled by man because they are not made by man” [ATL 79, p. 5], such as biological systems like memory, social systems, etc. The second shift takes an interest in an opposing way to the “building” of these worlds, hence the apparent paradox. By their prescriptive nature, teaching design (or Instructional Design) models are in fact essentially based on behaviorist-type epistemological approaches in order to “teach” the subject. Conversely, constructivist learning environments emphasize interactive environments that seek the personal experience of the subject, initially through collective activities, supported by the instructor and social group, and then through individual activities. The connectivist approach goes even further referring mainly to metacognitive, or even meta-metacognitive processes to which learners must access in order to learn through the network [SIE 05, BAT 77, p. 323]. In the latter cases, learning is no longer dependent on stricto sensu “teaching”. Formally, in constructivist, or even connectivist models, the learning process operates independently of cognitive situations constructed by the teacher or by the designer. At most, these environments require, on the part of the latter, “creation” of “real” environments using the context in which the study is appropriate [JON 09]. Since this is a matter of “artificially” designing “real” learning situations or simulating situations connected to a natural context, the teaching model design patterns gradually accommodated the constructivist beliefs and practices. The constructivist paradigm is now added to the founding paradigms of instructional design models after, first, behaviorist, then cognitivist [TRE 08]. From our point of view, one of the keys to the success of “the application of the complex systems theory to the modeling of new generation DLEs” lies in the connection between these two shifts. For Lemire [LEM 08, p. 38], building a world of knowledge first involves “weaving links between approaches to the modeling of reality and a general system”. This a way of explaining the changing world perceived as a system; to better understand and change with it, without wanting to force it. It is also pretending to take advantage of its evolution. It is learning more about the system in which we are making progress through the use of illuminating modeling or learning to better recognize ourselves as a cognitive system integrated into this encompassing system. By modeling such systems that surround us or live in us, our actions become metacognitive acts. It is by “applying a methodology to the systemic approach that models constructed from perceived realities become the representation of what is received as and by a general system” [LEM 08, p. 37]. From these considerations, the objective is clearly formulated as an invitation to model what characterizes a real world, conducive to the self-construction of knowledge.

164

Modeling of Next Generation Digital Learning Environments

4.2.2.2. Latest-generation DLEs: an “open” system Remember that what characterizes latest-generation DLEs, whose MOOCs are currently the best representatives, is primarily their openness to the world, as well as the consequences of this openness. The first ‘O’ (for Open) of the acronym MOOC ostensibly shows: online courses become open to all. As we recall from our book l’Appropriation sociale des MOOC en France [The social appropriation of MOOCs in France] [TRE 16a, p. 15], registration restrictions (place, time, age, degree) such as registration fees are dropped. Open means that everyone has access unlike traditional online teaching (distance learning), which requires registration with access conditions difficult to negotiate (degree, availability, staff, language proficiency, registration fees, etc.). This is one of the major features of MOOCs, which confirms and undoubtedly strengthens its originality. We can also immediately judge the consequences of this availability caused by the presence of the letter “M” for Massive in the acronym. The massive aspect of an MOOC is certainly what makes the biggest impression when you discover one. Most people questioned on this matter also begin by quoting some impressive numbers (hundreds of thousands of people) corresponding to the number of registered users. When you ask a teacher to describe his teaching methods, he immediately mentions about 30 students in front of him, and even more in the case of a university course. In an MOOC, there is not 30 or 50 registered, but 150,000 or even more. To our knowledge, the most registered for a single MOOC course has been 230,000 users to date. Imagine yourself in front of 230,000 students! This is substantially the size of the American population who came to listen to Martin Luther King Jr. (MLK) in his famous “I have a dream” speech in Washington in front of the Lincoln Memorial on 28 August 1963. Therefore, there exists an endless variety in the learner profiles, in their geographical and socio-cultural origins, in their prerequisites, but also in their ways of learning, in teaching methods, in assistive devices and assistance offered, etc. And since the complexity of a system is measured (even quantitatively) by its variety, there is no compunction in considering a latest-generation DLE as a complex system. As Tardieu et al. wrote [TAR 86] “The two notions that characterize complexity are the variety of elements and interactions between these elements”. Concerning interactions, we have seen that, within MOOCs, participants self-teach and motivate in an interactive and animated space [CRI 12], which, in view of the variety of systems constituents, increase the variety of interactions between these participants. Remember also that George Siemens20 is convinced that “formal education no longer represents the majority of our learning in that there are a large variety of ways to learn today, through communities of practice, personal networks and through work-related tasks”.

20 Personal translation of George Siemens’ quoted words [SIE 05].

The Digital Learning Environment in the Paradigm of Systemic Complexity Modeling

165

4.2.2.3. Unpredictable nature of DLEs It is also necessary to acknowledge the chaotic evolution of these environments that characterize complex systems. Their effects are unpredictable by nature. Going back to the exemplary case of MOOCs, nobody could in fact predict, in 2013, what they would become and least of all what they were supposed to bring to French society. Some of the most serious newspapers and media outlets also did not hesitate in predicting the end of traditional university education: “a model that could revolutionize the education system; with no need to follow prestigious educational courses”.21 And these may not just be the “visionary” qualities of Geneviève Fioraso22, then Secretary of Higher Education, who led the French government into the MOOC adventure in January 2014. This was rather the fear of missing out on any future profits, without being certain of receiving them at the end. The approach was, therefore, to not risk missing “a moving train”, while we knew very well its destination. Even today, nothing allows us to say with certainty to what extent this major innovation will change or not change the landscape of the French academic world and beyond. This unpredictability allows us to say that a latest-generation DLE can be studied as a complex system and not as an ordinary or “merely” complicated system (see section 4.1.1.2). Furthermore, complex (or chaotic) does not mean hazardous. In addition, there lies the interest in studying complex systems with systemic complexity modeling tools. As we have seen in section 4.1.1.3, between a completely predictable and a chaotic system (completely unpredictable) are complex systems (systems with average predictability). The whole point of systemic complexity modeling is unveiled: the emergence of certain phenomena can be anticipated by modeling with a certain probability of accuracy. Therefore, we can consider not only all sorts of predictions that will be useful to developers and users of DLE, but also policies that do not cease to request enlightening reports on these social issues. This is why we devote a whole section to this topic (section 4.2.5). 4.2.2.4. Consequence of the limits of analytical modeling and disjunctive logic Disjunctive logic specific to analytical modeling (AM), as opposed to individual conjunctive modeling specific to systemic modeling (SM), aims to independently study the constituents of a system: “the breakdown results must be definitively distinguished AND separated” [LEM 99, p. 32]. In fact, it does not allow for 21 … said in an hour of prime time, by reporter Claire Chazal on the television program Le JT week-end on TF1 at the start of the 2013 academic year! 22 Interviewed on 8 March 2014 during the program 01 Business.

166

Modeling of Next Generation Digital Learning Environments

“the realization of phenomena that we perceive in and by their complex conjunctions” [LEM 99, p. 33]. For example, let us take the example of the energy and work concepts in physical sciences. Applied to a body under gravity in space (for example, a rock resting on our hand a meter from the ground), we define the potential energy of this body by the amount of the work (in the physical sense of the term, that is, force times displacement) it is “potentially” capable of producing during its fall (if we drop the rock, of course). Therefore, it is easy to find the value of its potential energy: m.g.z (m.g being the weight of mass (m) in a gravitational field (g) and z the distance from the rock to the ground). We see through this example that it is impossible to define the concept of energy without talking about work and vice versa. These concepts cannot understand each other and can only observe each other in their conjunction. Many other examples can be cited to illustrate this need to conjoin rather than separate. In the case that concerns us, we ask the question to know what is the point of separating the “subject” and the “tool” or the “subject” and its “community”, just as Engestrom did in his model. Can these system constituents not also be seen in their conjunction? Bogdanov23 [BOG 81, p. 63] cited by Le Moigne [LEM 99, p. 33] considers the act of joining as fundamental, and separation as derivative; “the first is direct, the second is the result: all begins with the act of joining”. Let us look back for a moment at the anthropocentric approach of the Rabardel techniques [RAB 95] seen in section 1.2.4 in Chapter 1. Is it not also intended to transcend this mind/matter dualism to approach the cognitive development of the subject by placing it in its social and material environment? Piaget and Vygotsky also worked together to overcome a Cartesian dualistic vision (mind/matter) by locating the thought mechanism and its development at the crossroads of an outside/inside, middle/subject loop. According to Bogdanov: “The two conceivable actions of man in nature, whether practical or cognitive actions, are to join and to separate. […] Join his body or attention to the system to be considered; apply a conscious effort, which is still a conjunction. It is a similar situation for cognition: no distinction is possible without a preliminary comparison, which is still 23 Alexander Bogdanov “published in Moscow, between 1913 and 1920, Essays in Tektology: The General Science of Organization, which today, seems to be recognized as the first full general systemic treaty! A few English translations of this untraceable text have become available since 1980 and it is reported that the Soviet authorities have recently begun to allow its review!” [LEM 94, p. 10].

The Digital Learning Environment in the Paradigm of Systemic Complexity Modeling

167

an act of conjunction. In this case, disjunction is also secondary.” [BOG 81, p. 63] Bateson [BAT 77, p. 275] reinforces the idea of inseparability of the “subject” and the “tool”. To do this, he takes the example of a blind man and his cane (tool) used to help guide him. “The cane is simply a way [he says] in which transformed differences are transmitted, in a way that the cutting of this channel removes a part of the systemic circuit that determines the possibility of locomotion for the blind man”. Michel Serres [SER 15] also shares that view in a French metaphor: “la tête de Saint Denis” (English translation: the head of Saint Denis). He made an analogy between “our head” and “our computer”. We have, he says, “our head in our hands” since we are more than ever in an externalization logic of our cognitive functions in the direction of tools capable of memorizing and processing information instead of us. These tools become part of ourselves (our head is in our hands) and therefore inseparable: the subject and the tool are one. But what should we then think of all these models that assert systemic nature and that, as we ourselves have done, represent only the components of a system, connected through relationships? Le Moigne provides an explanation: “Between 1960 and 1980, many treaties outlined, under the name Systems Analysis (plural matter), a methodology of analytical modeling hidden under a systemic jargon that was able to, at the time, allude to the fact that before we agreed that this systems analysis only attributed to closed and complicated but not complex systems analysis modeling. But since 1975, we can consider that the systemic and fundamental engineering sciences have begun to emerge from their initial conceptualization and epistemological maturation phase, to enter into a strictly instrumental and methodological phase: it will then be possible to present systems modeling in its originality, without coating it in the finery of analytical modeling in order to improve its respectability.” [LEM 99, p. 7] It is therefore, unbeknownst to users, very often “through scientific willingness” as the author emphasized [LEM 94, p. 8], “that the use of ‘analysis system’, ‘system approach’, and ‘application of systems theory’” concepts has almost always led them to practice analytical modeling that only had systemic in its name. This is the lack of adaption to these internal epistemic critical exercises that should be familiar to any scientific and technical activity. However, it is not difficult to first ask oneself “what does that do, in what sense, for what, becoming what?” rather than wondering firstly “what is it made from?”.

168

Modeling of Next Generation Digital Learning Environments

4.2.2.5. For a functional and non-organic approach To model a DLE, we have taken the position of only supporting a secondary attention to its structure to focus on rather “what it does”. We will represent each of its actions or complexes of actions by “a black box or the symbolic processor that takes action into account” [LEM 99, p. 46]. We will see in Chapters 5 and 6 that this decision occurs in practice when entering modeling: not with objects but with functions (“functional diagrams” in OMT and “use cases” in UML). It is also possible to verify that this choice is confirmed at the level of the necessary simplifications that we will have to make regarding the data to be processed. The use of multivariate descriptive statistical techniques aiming to intelligently conjoin our factorial axis data and producing representative classes (clusters) of specific populations (which facilitate dendrograms) once again demonstrates the adoption of this conjunctive logic. 4.2.3. Complexity theory: at the heart of the connectivist learning design We have said and shown repeatedly: when a DLE model is subjected to too many significant variations in initial conditions (Chapter 3), the adjustment of the intended model to represent (learning, educational, training, etc.) becomes very difficult, if not impossible. That is why we have proposed changing the paradigm, choosing a systemic modeling of complexity that offers a more suitable framework. Here we will address this issue from the perspective of modeling forms of networked learning. Contrary to traditional distance education, which seeks to reproduce within DLEs from real, significant situations susceptible to promoting apprenticeships, MOOC professionals (designers, researchers, teachers, etc.) or more generally OpenCourseWare (OCW) implicitly poses the (connectivist?) assumption that the meaning of the activities (pre)exist: the learner’s challenge is to recognize the reoccurrences (patterns) that appear hidden [SIE 05], resulting in their decision-making process. If the conditions involving their decision-making process change, this decision will no longer necessarily be correct. It was at the moment when they took it but will no longer be after this change [SIE 05]. Connectivism is based on the idea that the strategies that we deploy to learn rest on unstable models and contexts and whose foundations are moving. Our ability to recognize and adjust our strategies to changes in context is an essential task of learning. Bateson [BAT 77, p. 307] wonderfully formalized this change in how to learn according to the signals emitted by the context, themselves dependent on the signals emitted by the individual. He establishes three levels of learning all punctuated by a change of one kind or another. “Change can define learning”, he says.

The Digital Learning Environment in the Paradigm of Systemic Complexity Modeling

169

“The word ‘learning’ undoubtedly indicates one kind of change or another but it is very difficult to say what ‘kind’ of change it is […]. Change involves a process. But the processes themselves are exposed to change.” [BAT 77, p. 303] We have already posed the assumption [TRE 05, p. 66] that learning within a DLE must be “highly” facilitated by the acquisition of a “Level III Batesonian”. Let us look back for a moment at these three fundamental levels, which have not diminished in learning contexts, holding our attention here. Bateson [BAT 77, p. 307] develops the idea that all learning is to some extent “stochastic” (it contains sequences of trial and error). He introduced “Level 1” learning corresponding to a change in the response specificity through a correction of errors of choice within a set of possibilities. In a Level 1 training context, the individual learns and therefore corrects his actions and his choices and corrects his mistakes within a set of possibilities (a given conceptual framework). The learner researches contexts or sequences of one particular kind rather than another. Learning to learn (“Level II Batesonian”) becomes the acquisition of abstract thought patterns. The subject has acquired a certain talent, a certain skill in Level I. Level II is the change in the Level I learning process: it has learned how to learn. It will thus react better and faster. Level II will be adaptive if the subject is not wrong in the assessment of the contingency model it encounters. Then, Level II sets up, in time, a certain cognitive comfort, a mental habit. The effort is less strained. An enhancement at Level II allows an economy of thought. However, if the educational environment changes, the conceptual framework also changes, changing the set of choices. This habit to act in a given context is called into question, sometimes even painfully, because it is a matter of getting rid of a habit that worked well before the change of environment. This change in Level II, allowing access to new routines or the replacement of an existing routine, characterizes “Level III”. When it comes to a replacement, it often follows generated contradictions at Level II, contradictions all the greater because mental habits are rooted since they last over time. To avoid the “painful effect” of such a replacement, it is possible to limit Level II learning. The replacement of the habit that is ineffective is made less brutally by the perception of context indicators. “This restructuring of the beliefs of the considered entity” [ANC 05, p. 20] is a change in the learning process and, therefore, learning itself. It would be “learning to learn” and this by recognizing meta-metamessages.

170

Modeling of Next Generation Digital Learning Environments

Thus, there would be new capacities that learners would develop, or would be angling to develop, in a latest-generation DLE; this is at least the assumption that we formulated. The knowledge acquired through this bias is then “dispersed” into the four corners of digital networks. It then becomes easier to navigate this complexity and to reassemble the required knowledge once such metacognitive skills have been acquired. Personal knowledge has in fact become, little by little, a knowledge network that feeds on organizations and institutions, which in turn feed on the network that continues to feed the person learning [SIE 05]. “This cycle of knowledge development (personal to the network for organization) helps learners to remain up-to-date in their field through connections they have formed.” [SIE 05] “[For Siemens,] connectivism is the integration of principles explored by ‘chaos, network, and complexity and self-organization theories’. Learning is a process that occurs within ‘nebulous environments’ with shifting core elements, not entirely under the control of the individual.” [SIE 05] “[Thus,] connectivism is part of several phenomena specific to current professional activities: chaos (everything is related), complexity, networks and self-organization. What forms the heart of the theory of connectivism is the role of relationships and flows between individuals and computers that speed them up and not only the knowledge content.” [CRI 12] 4.2.4. Conceptual analogy between the SM of complexity and SM of activity We want to show quickly here that there is a conceptual analogy between these two theories, which reinforces the idea of applying complexity and its modeling to instrumented systems activity analysis just like latest-generation DLEs. If we observe the models from the theories of activity (see section 1.2.5 in Chapter 1) carefully, from Vygotsky [VYG 97], Léontiev [LÉO 76], Engestrom [ENG 87], etc., we find that they all deal with human activity, like a system that acts and is transformed by acting. This concretely translates through the realization of a succession of functional operations grouped into a set of actions forming activities to accomplish tasks.

The Digital Learning Environment in the Paradigm of Systemic Complexity Modeling

171

Let us cite a short excerpt from the content of a course at Ecole des Mines de Saint-Etienne on complex system modeling to compare his approach to the previous: “The system is understood at the macroscopic level through a group of ‘processes’ competing at work and the microscopic level by all of the ‘basic actions’ distributed in time and space. These functional operations can be grouped into sets of ‘actions’ forming ‘activities’ to accomplish ‘tasks’; three fundamental concepts from general to atomic are: ‘process, activity, operation’”. [AUG 13, p. 6] Is it necessary to emphasize (concerning associations) the possibility of establishing this formulation of activity and that of Léontiev who refuses activity in one set of composed operational actions? Each refusal provides three noticeably identical hierarchic levels both within the same activity, even if it is true that the boundaries between these levels sometimes remain porous or moving in each case… We are aware that it is not possible to generalize from a single example. However, it would be quite easy to find others and show that the two approaches are noticeably equivalent. Dir and Simonian [DIR 16, paragraph 4], reveal, for example, that “activity is not reduced to action: the first requires a ‘holistic approach’ of cultural mediations involving a collective (and especially) a cooperative dimension while the second focuses on one individual”; these are also characteristics specific to systemic complexity modeling. Better yet, for Engestrom, quoted by Simonian [SIM 10, p. 58]24, “activity involves individual actions, it is in no way reduced to the sum of individual actions”. Is this not another way to introduce the emergence concept specific to systemic complexity modeling that can be roughly summarized by the adage: “Sometimes everything has more possibilities than the sole sum of its parts”. Can a property indeed qualify as emerging if it is “derived” from more fundamental properties while remaining “new” or “irreducible” to them25? It seems to us that the common denominator between these two approaches is precisely this “activity” to which we will readily substitute the most unifying concept of Praxis, that has no other purpose than the achievement of man (art, politics, music, knowledge), his own transformation, through the transformation of the natural environment and social relationships.

24 See also: Dir and Simonian [DIR 16]. 25 “Emergence”, Wikipedia (2018, https://en.wikipedia.org/wiki/Emergence.

updated

18

September),

available

at:

172

Modeling of Next Generation Digital Learning Environments

This achievement is nothing more than a phenomenon emerging, such as the mind, thought or even knowledge. To our knowledge, the model established by Engestrom [ENG 87] is one of the latest major developments in the theoretical framework of the theory of activity. 4.2.5. Modeling these environments with suitable languages 4.2.5.1. A common interpretation of the concept of “processes” In sections 2.4.1 and 2.4.2 in Chapter 2, we mentioned the evolution of successive instructional design and training models (EML) aimed to respectively create scenarios for learning situations and training mechanisms. “With the evolution of learning theories and the rise of ICT in education, the scenario has become an indispensable tool to guide the work of the teacher and the learner. This tool has a central place in computerized environments.” [HEN 07, p. 16] Whether they are “educational” or “training” engineering, these models have the common character that they are both “exploratory tools that help developers to define the outlines of an ITS; these languages also allow the various actors to exchange view points within a development team to achieve a viable solution” [NOD 07a, p. 85]. We recognize in Oubahssi and Grandbastien’s canonical model [OUB 07], explained in section 2.4.2 in Chapter 2 (which refers to the ISO 9000 standard), a striking conceptual analogy with the systemic complexity model, based on the concepts of “processes” and “processors”. At the ISO 9000 level, a process is represented by a processor and defined “as a set of correlated or interactive activities that transforms entry elements into output elements” [OUB 07], while in systems complexity modeling, a process can be represented by a system of multiple actions or by a process that can be a tangle of processes that will systematically be represented by the black box. It is also defined “by its exercise and result (an ‘implex’ therefore)” [LEM 99, p. 46].

Figure 4.7. ISO process model (ISO 9000)

The Digital Learning Environment in the Paradigm of Systemic Complexity Modeling

173

Figure 4.8. SM processor model

For Oubahssi and Grandbastien: “The input elements of a process usually form output elements of one or several other processes. Each process is described, at a first level, by its input and output data (process result). Several processes can occur in the lifecycle of a product. The processes involved in the lifecycle of a product describe the set of related means and activities that transform incoming elements into outgoing elements. These means may include personnel, finance, facilities, equipment, techniques and methods. Figure 4.7 represents a process in a simple way.” [OUB 07] For Le Moigne: “It was agreed to designate a processor by Pr, the symbol of the black box by which ‘a process’ is represented (Figure 4.8). Each processor (Pr) can be characterized in each period (ti) by the values assigned to its ‘inputs’ and ‘outputs’ (inputs and outputs are outlined by vectors, whose components are designated by the modeler referring to his modeling project modeling).” [LEM 99, p. 48] “The preconceived identification of N processors and input and output couples by which one characterizes them is not given by the problem that the modeler considers” [LEM 99, p. 54]. This will therefore be a part of the projects that we will address in the modeling design process. The latter will be built by successive iterations “between projects and symbolic representations that the modeler has made” [LEM 99, p. 54]. To conclude on this point, we acknowledge that the central process concept, which lies at the heart of systemic complexity modeling, does not come into conflict with that of traditional instructional design (or training); no antinomic character is initially detected.

174

Modeling of Next Generation Digital Learning Environments

4.2.5.2. Modeling complexity by object-oriented modeling Beyond the developed theoretical aspects aimed at showing the interest in applying the complexity theory to the modeling of latest-generation DLEs, we wanted to offer a concrete continuation to this project. We intended to implement the recommended approach by developing models for research. We had to adopt a modeling language, but obviously not just any language: a language in line with the epistemological conceptions of SM, but that was also capable of meeting the challenges that we set. After some reflection, we turned first to a “technical” modeling and then secondly to a modeling language specific to this technique. It related to “object-oriented” modeling, and especially Rumbaugh26 et al’s OMT (Object Modeling Technique) [RUM 95]. We have chosen this because it follows the main “principles” of the complex systems theory by offering both a functional representation of the system (the system “created”) and a dynamic representation (the system “becomes”), synchronously, without established order (the system “is created and becomes so by doing”). We will justify this important choice (that of the OMT modeling “technique”) point by point before clarifying the object-oriented modeling language that we have retained. With the choice of modeling language being a direct consequence of the “technique”27 used, it will be easy to justify, as this language was in part developed by the same author as the OMT “technique”, James Rumbaugh. The first advantage of object-oriented modeling is that it is visual and allows for modeling a DLE not only to analyze it but also to design it. Therefore, it is a sort of an “all-in-one” product. Moreover, object-oriented modeling is also based on an exploratory and participatory approach (that we were absolutely looking for), allowing the modeler to recognize user needs very early by continuously exchanging viewpoints with them (iterative process) using an understandable medium for all (intelligibility character specific to SM). One of the features of object-oriented modeling, which could be initially seen as a barrier, is that it is often presented as an approach that opposes “functional decomposition modeling”, which can be annoying concerning the basic concept of system modeling which is the action. “SM leaves behind the question, ‘What does 26 James Rumbaugh is the creator of the “object-oriented modeling” language. He is also one of the three founding fathers of UML, with Ivar Jacobson and Grady Booch (source: Wikipedia). 27 In reality, this technique is also one of the first modeling languages. However, as it was called a technique by its designer, we have decided to keep this wording in order to not unnecessarily complicate things.

The Digital Learning Environment in the Paradigm of Systemic Complexity Modeling

175

that do?’: what are ‘the functions’ and transformations, or ensured operations or operations to be ensured?” [LEM 99, p. 46]. However, looking more closely, it is not the functional aspect that is questioned in the object-oriented approach, but rather artificial decomposition of the overall function of the system in more simple functions to understand or program. In other words, object-oriented modeling does not enter into the project in terms of its functions, but it opposes any hierarchical and artificial segmentation, as indeed does SM with concerning AM. By contrast, object-oriented modeling recommends an entrance into the modeling project, precisely asking, “what is its purpose?” or “what will be its purpose?” (what needs it answers or will have to answer). This is why this approach advocates modeling the needs before anything else, making an inventory of the functions, organizing the needs between functions so as to show their interrelationships. We also needed to clarify the “object” concept that could be perceived by SM supporters as unsuited to a systemic view of modeling. To illustrate this reluctance, let us cite a passage from Le Moigne [LEM 99, p. 46], who willingly claims to be a provocateur with regards to our argument: “The basic concept of SM is not the object, […] but the action [..]”. So why choose an object-oriented modeling technique? On the SM side, Simon, 1969, quoted by Lecas [LEC 06, paragraph 52], insists on the tangible reality of “objects” produced by the engineering sciences: the computer, operating accounts, and forecast budget are objects like others and therefore objects of scientific studies. However, he indicates immediately afterwards that it will benefit from the fact that “these objects are defined as the result of an identifiable project” to reverse the knowledge method: instead of dissecting the computer, the budget to discover the natural laws that govern their behavior, he proposes modeling the cognitive process; the process developed by the project that defines and explains possible and intentional (and no longer necessary and determined) behaviors. This is precisely the sense that Le Moigne gives to the basic concept of SM when he leaves the SM “object” to favor the “action”. In object-oriented modeling, the “purpose” is anything but a component of the system such as the one representing analytical modeling (AM). It is a concept that assembles (conjunctive approach) actions and information in the same entity (which may be artifactual, natural or man-made) rather than separate them. This approach is very different to a conventional analytical approach in which data structures and behavior are often only weakly related. Before closing this argument concerning the adaptability terms of object-oriented modeling to SM, notice that the OMT approach follows remarkably well the fundamental principles of the iterative and incremental approach advocated by SM, characterized in particular by Boulding’s seven levels of complexification

176

Modeling of Next Generation Digital Learning Environments

[BOU 04a]. For example, in OMT modeling, as soon as actions that require “decisions” appear, their treatments are never immediately represented in the model. They will be in the later stages as soon as the decision-making system has been constructed. And to remember these actions (awaiting decisions), they are represented, as suggested by the authors [RUM 95, p. 181], by sporadic traits (see Figure 5.20). 4.2.5.3. Using UML as a modeling language Heavily based on the paradigm of “object-oriented” modeling, to us, UML (Unified Modeling Language) seemed to perfectly suit the needs expressed so far. Therefore, we quickly adopted it, especially as James Rumbaugh, OMT founder, is also “one of the founding fathers of UML (along with Grady Booch, founder of the Booch language and Ivar Jacobson, founder of the OOSE language)”28. Unified means that UML results in the merging of these previous languages, all “object-oriented”. UML is now a standard adopted by the Object Management Group (OMG). In detail, it is a graphical modeling language which, from pictograms, allows you to view a functioning system. It is not a method, since each is free to use the diagrams deemed to be good. No other approach is therefore advocated. Each uses the diagram types it desires, in the order it desires; it is sufficient that the diagrams implemented are consistent with each other before eventually moving to a development (to design or simulate). It is indeed possible, in a second instance, to move from the system “representation” (the model) to “executable”. Used just as much by humans as by machines, the graphical representations have sufficient formal qualities to be automatically translated into a skeleton source code (development in Java, Python, C++, etc.). UML officially came about in 1994, while UML version 2.0 dates back to 2005. This was a major release bringing radical innovations and widely extending the UML scope. The last official version according to the OMG (Object Management Group)29 is UML 2.5 (March 2015). Here is a list of some UML modeling software available on the market30: StarUml, ArgoUml, BoUml, PowerDesigner, etc. On our side, we have used ArgoUml 0.34 for a few brief illustrations in Chapter 5 and StarUml 2.7.0 for a 28 “Object modeling technique”, Wikipedia (2018, updated 28 July), available at: https://en.wikipedia.org/wiki/Object-modeling_technique. 29 OCL 2.5 specifications are available for download from the following link: http://www.omg.org/spec/UML/2.5/. 30 A comparison chart is available (in French) at the following address: https://manurenaux.wp. mines-Telecom.fr/2013/09/26/Comparatif-vous-cherchez-UNmodeleur-UML/.

The Digital Learning Environment in the Paradigm of Systemic Complexity Modeling

177

more “formal” representation, especially from the perspective of a simulation of the system described in Chapter 6 (practical component). UML offers a couple of diagrams (varying in number depending on the versions) grouped into diagram families: “static”, “behavior”, “interaction or dynamic”. We will obviously not review all of these diagrams. It should be noted that entry into modeling is usually done by “use case” diagrams (representing the performing system) rather than static diagrams (representing the existing system), which is once again in full compliance with the recommendations of SM. Concerning complex systems theory, it begins with the identification of the projects or intentions of the modeling system that constitute the model. It is from these projects that SM will propose to initiate the model design process. UML follows this logic; it is only after the “use cases” diagrams, obtained and constructed using data from interviews with users, tracking tools, dashboards and simulations (see section 6.2 in Chapter 6), that other charts will be deployed. Moreover, UML is a semi-formal description language that has a dual advantage: it can be presented in the form of a set of quickly comprehensible diagrams and has the ability to be transformed (using software) into programming languages (C++, Java, Python, XML, etc.)31: we therefore felt that it was well suited to the formalization of our problem. In fact, different domains have created transformation software tools going from UML schematic models to formal applications (most often a computer application) through successive refinements. Therefore, to us it seems that such tools could be used to serve a team of teachers wishing to focus on a set of UML diagrams to formalize learning situations in a DLE, the tool to transform these visual diagrams via an XML formalization in a compliable structure (for example, a simulation or an implementation on a platform). This means of course that the education team has a sufficient command of the UML language, which is not the most difficult to master. Nodenot, Laforcade and Le Pallec re-establish, for example, mechanisms offered by UML to provide a visual language (CPM32), useful in distance education:

31 For transformation, it is possible to use the XSL language that allows for the joining of a code (PHP, HTML or anther) to an XML element, therefore a UML element. In other words, for a UML forum element, XSL can be converted into a code for a forum that functions (in HTML, PHP with database, for example). 32 “Preliminary work was conducted in order to offer a language, CPM (Cooperative Problem-based learning Metamodel), exploiting the graphical richness, proposed by UML (Unified Modeling Language). CPM is therefore constructed in the form of a UML profile. It is positioned at the design phase level above EMLs (Educational Modeling Language)” [LAF 05a, p. 213].

178

Modeling of Next Generation Digital Learning Environments

“To describe the sequencing of activities, detail the resources exchanged between actors during these activities, identify their performance condition based on the values taken by resources or results of previous activities, etc. ‘UML diagrams’ are also used by these languages to express most notably the lifecycle of resources and activities as well as the significant events of the concepts described. […] The design process begins with use cases diagram production. Each diagram is then refined by other diagrams to use boxes and then by one or more ‘activity charts’ to describe what is happening within an activity defined at the top level. In the case of ‘collaborative activities’, each corridor ‘of the activity diagram’ allows for the determination of specific tasks conducted by an actor as well as the sequencing of these tasks. At each of the abstraction levels, ‘class diagrams’ and ‘state-transitions diagrams’ allow for the characterization/specification of modeling elements highlighted in other diagrams.” [NOD 07b, p. 92]. 4.2.6. Enrolling in a forward-looking approach The study of emerging properties that characterizes any complex system (see 4.1.1.5) allowed us to recall that a cause, however small it may be, can have a significant impact on complex system behavior. As the words of Henri Poincaré highlight (1854–1912), quoted by Rocet when he laid the foundations of the chaos theory by studying the three-body problem: “If we knew exactly the situation of the universe and the laws of nature at the initial moment, we could exactly predict the situation of this same universe at a later moment. But, even though the laws of nature would not have been a secret to us, we could only approximately recognize the situation. If this allows us to predict a future situation with the same approximation, that’s all that we need, we say that the phenomenon was expected, it is governed by laws; but it isn’t always the case since it may happen that small differences to initial conditions lead to very big differences in the final phenomena; a small error on the first would be a huge mistake on the last. The prediction becomes impossible and we have the coincidental phenomenon.” [ROC 13a] But how then do we predict events without knowing the rules that tell us how the system will change, even approximately? To do this, we explained in section 4.1.1.3 that we first had to situate ourselves in a system whose level of complexity lies between two extremes: a so-called algorithmic complexity (that is, that the rules that

The Digital Learning Environment in the Paradigm of Systemic Complexity Modeling

179

tell us how the system will change are known to man because they are built by him) and a chaotic complexity, whose rules are not absolutely known to man. In a more formal manner, we are interested in systems whose complexity (measured by the amount of system information in the Shannonian sense of the term) is intermediate to average predictability: a mix of knowing and ignorance. And this is the case of DLEs precisely located between these two extremes that are the “crystal” and the “smoke”, as Atlan would say [ATL 79]. Complexity is not only the result of “varying degree of elements that constitute it” [ATL 79, p. 45], but also “the result of the large number of states that a system can take on, taking into account the interrelations that can build its parts. These states are not predictable by deterministic analysis since the system is open. Many of these states are emerging” [ROC 13b]. Therefore, to enroll in a forwardlooking approach, proceed with the construction of static models that will represent the variety of system components and the construction of dynamic models that will represent (possible) different states of its constituents. Emerging phenomena will be revealed by system modeling and simulation. This is a very time-consuming activity, because if the emerging properties cannot be explained from characteristics of the parts taken in isolation, the latter must nevertheless be known and modeled. The properties of the action complex only appear as “new” or “emerging” to this condition. The issue of a forward-looking approach is now vital for our society where we live in a strong, complex world. Whether this approach is conducted for political, economic, social or simply cognitive purposes, it will ultimately lead us to “adopt precaution and ecology principles, a policy for the future taking into account our fragility, the limitations of our resources and vital balances” [ZIN 03]. Our book on the appropriation of MOOCs in France [TRE 16a] shows the vital need to have some of these forward-looking elements to implement a policy in terms of digital development at the national or even international level, whether it be by taking on the adventure of MOOCs or through equipment plans in schools and colleges (iPads, for example). A forward-looking approach aims to prepare us today for tomorrow. It is not a case of predicting the future (which would be a sort of divination), but of developing plausible scenarios in their perceptions based on the analysis of available data (surveys, heavy trends, emergence phenomena).

5 Modeling a DLE Perceived as a Complex System

We have just explained in the preceding chapter that we are choosing to study latest generation DLEs, referring to the theory of complex system modelling, because we perceive them as complex phenomena (for example, an MOOC). Now, let us look at the different stages of this model and its instrumentation. According to the systemic modeling theory, modeling a complex system first consists of modeling a synchronic action system1 (that functions), a diachronic action system 2 (transforming while functioning), a teleological action system (having a purpose, a goal) and recursive action system (automation) in an active environment. Systemic modeling also passes by respect of a conjunctive logic which aims to join and not separate the active environment and project (or teleology) concepts on the one hand, and those of synchronic functioning (doing) and diachronic transformation (becoming) on the other. The cybernetic procedure characterizes the conjunction of the first two concepts; the structuralist procedure, the conjunction of the last two. The general system (GS) concept results in the conjunction of these two modeling procedure support concepts, namely the cybernetic and structuralist procedures.

1 Designates phenomena or processes occurring at the same time, not successive. (Source: Le Petit Lexique des termes de la complexité from the “Réseau Intelligence de la complexité”, available (in French) at: http://www.intelligence-complexite.org/fr/documents/lexique-determes-de-la-complexite.html.) 2 Adjective meaning something that extends, evolves or changes over time (as opposed to being synchronic). Same source as above. Modeling of Next Generation Digital Learning Environments: Complex Systems Theory, First Edition. Marc Trestini. © ISTE Ltd 2018. Published by ISTE Ltd and John Wiley & Sons, Inc.

182

Modeling of Next Generation Digital Learning Environments

Figure 5.1. Systemic conjunction [LEM 99, p. 40]

Systemic conjunction therefore proposes “to keep inseparable ‘the operation and transformation of a phenomenon, the active environments’ in which it is carried and ‘projects’ against which it is identifiable” [LEM 99, p. 40]. “Any complex system can therefore be represented by ‘a system of multiple actions’ or by ‘a process’ that can be a tangle of processes. This can be represented by the designation of identifiable ‘functions’ that it exercises or may exercise” [LEM 99, p. 48].

Figure 5.2. Process identification

Systemic modeling (SM) does not begin to represent things, objects, individuals, bodies, as is done in analytical modeling (AM), but an action (characterized by a process) that can be a tangle of actions (or a complex of actions) that one will systematically represent by the black box (or a symbolic processor)3 that accounts for this action or a complex of actions. This is the basic concept of systemic 3 In terms of modeling technique, an active processor is symbolized by the black box.

Modeling a DLE Perceived as a Complex System

183

modeling (SM) 4 other than from what “the system performs”, in other words, operations and transformations or ensured operations or those to be ensured. However, “identifying preconceived N processors and input and output couples by which one characterizes them is not given by the problem that the modeler considers” [LEM 99, p. 54]. Therefore, starting from the projects that we will address, this will be the model design process. The latter will be built by successive iterations “between projects and symbolic representations that the modeler constructs” [LEM 99, p. 54]. 5.1. Finalized and recursive processes in an active environment 5.1.1. Identifiable finalized processes (functions) Initially, the modeler has a basic perception of the world to model. This first stage of the systemic modeling of complexity corresponds to the first level of an archetype model for the connection of a nine-level complex system as described in section 5.1.2.2. Its perception, at first syncretic, of the system or phenomenon to model does not allow it to see the details. It then determines its outline, its borders5 with the outside world and examines its global function that it will later set out. The purpose of the project will be described in its dynamic dimension. Without going into the details of this description that strongly depends on the type of project to model6, remember that the main function of a DLE is to allow the learner to acquire knowledge and develop skills through the use of a digital environment. Variations to this general description are justified from the moment the DLE is considered or not considered by the modeler as a new generation DLE. If so, it must assume its complexity and describe what characterizes its function. In the case of an MOOC, for example, modeling will be necessarily different from that of a traditional distance education mechanism. For example, the project would be to “educate online all those registered for this course, without preconditions (age, diploma, pre-requisites, registration fees, enrollment, language skills used, etc.)”. A first representation could be shown in Figure 5.3.

4 Remember that AM acts differently. It starts from the system structure, the identification of bodies, system components, in other words, it assigns itself the mission of explaining the system composition, answering the question: “what is it made of?” (“what are the components?”). 5 This could be “this world” in the case of an MOOC, for example. 6 In ITyPA creation, for example, only the project was truly significant.

184

Modeling of Next Generation Digital Learning Environments

Figure 5.3. DLE represented by a black box

It is at this stage that the modeler can register the system (the process) in its environment; the project becomes identifiable and differentiable from its environment. The latter is also composed of different actors7. “An actor represents a coherent set of roles played by the entities that interact with the system. These entities can be human users or other subsystems” [MAM 11, p. 8]. Among these actors, we will examine the primary actors from the secondary actors. The primary actors act directly on the system. In the case of a DLE, this mainly relates to the learner subjects, teachers, tutors, training managers, computer scientists, programmers, designers and developers. They all need to use the system, the first group to learn, the second to teach and the last to maintain the system in good working order. Secondary actors do not have a direct use of the DLE. They can nevertheless be called upon or consulted by the primary actors in the system, either to exchange information or to meet specific needs. These are usually actors outside the system considered as service providers (for example, Moodle in the first MOOC EFAN8) or simply partners (for example, those of Coursera9). YouTube and Dailymotion are other examples of secondary actors, both offering hosting of videos to MOOCs. This is precisely the case of Dailymotion who hosts FUN (France Université Numérique) MOOC videos. 7 To use a generic term specific to modeling a language. 8 MOOC EFAN: Enseigner et former avec le numérique (digitized teaching and training). 9 COURSERA: One of the first American MOOCs (Stanford University, California).

Modeling a DLE Perceived as a Complex System

185

Some actors are of the “human” type, and others not. To differentiate between them, some modeling languages represent human actors by stick figures. Their respective roles in the system are indicated in Figure 5.4.

Figure 5.4. A representation of a human type actor

To represent a “non-human” type actor, the modeler can either choose a representation among those proposed by current modeling languages or choose its own representation. In the first case, the representations may vary. This goes from the stick figure non-human actor with a large black bar on its head to the non-human actor represented by the same stick figure, but with an indication of the stereotype, through to a simple box with an indication of the stereotype (see Figure 5.5).

Figure 5.5. A representation of a human type actor

Continuing with the example of an MOOC for illustration purposes, the latter would be made up of human and non-human actors who interact with the system. DLE users who need this environment to learn would naturally be a part of the human actors. Video webhosts belong to the category of non-human actors. The black box becomes the opportunity to learn for human users.

186

Modeling of Next Generation Digital Learning Environments

Figure 5.6. DLE placed in its environment

The description of nested processes or complex actions ensues from this activity, which will gradually fill the black box, which represents both the system and its overall project. “with the families of projects, we will associate subsystem hypotheses that one seeks to connect... referring to the overall project of the modelling system (and not to the hypothetical nature of things). Then departing from the levels, we will attempt to put them together in a processor system, as a composer deliberately looking to compose a musical system with the help of a symbolic representation system. Modeling a complex system will organize a series of iterations between the projects and symbolic representations that the modeler has contracted” [LEM 99, p. 54]. This procedure is what Le Moigne calls systemography10. This set of multiple actions or processes identified by the modeler can take place inside of a concept map, created in order to represent them (see Figures 5.7–5.11). It is advisable to clarify the perceived traits of the phenomenon to model everything by completing the map.

10 The term used by Le Moigne [LEM 99, p. 27] by analogy with photography: “procedure by which we build models (clichés) of contrasting forms of phenomena that we see through a lens”.

Modeling a DLE Perceived as a Complex System

187

“It should be noted that these taxonomies are based on the needs of a specific world. They are often limited, considering real-life experiences of the subject(s)-explorer(s). They are derived from customized investigations of the world to model; then, they are consolidated by the consensus approach undertaken by the project team, when there is one, and mediators, bearers of scholarly knowledge. The team agrees to share objects, terms and definitions to come to a relevant judgment in relation to the components of the world which should be taken into account throughout the project to complete. This worldly description is first based on the intersubjectivity of the people involved: they support their judgment through the use of real-life experiences, or they justify appropriateness taking position in a subjective way. This is the description of a complex system and layout of its components in various models that will result from this cognitive activity” [LEM 08, p. 75].

Figure 5.7. An example of a blank process system concept map

These representations include structural matrices, arrangements, tree diagrams, conceptual maps or any other development that can be used in the representation of projects, actions, complex actions, functions (processors) or simply relationships that unite them.

188

Modeling of Next Generation Digital Learning Environments

To facilitate the representation of these processors, we agree to indicate them by “Pr”, the processor symbolizing the “black box”, and by “ti”, the period during which the values of its inputs and outputs are assigned. The processes characterizing the active phenomenon are now perceived in their actions (acting within the system). Peraya [PER 03], Meunier and Peraya [MEU 04], Charlier Deschryver and Peraya [CHA 06] were interested in the “approach by constituent functions of any publicized educational environment” without (to our knowledge) having an explicit reference to the theory of complex system modeling. Today, the analysis framework relative to these constituent functions of any publicized educational environment constitutes a frame of reference. Based on different taxonomies that scientific literature proposes11, these authors retain eight functions, namely (1) awareness or highlighting of “signs of the presence of distance actors” [WEI 03, JAC 06]; (2) social interaction which includes what is conventionally referred to as collaboration, communication and exchanges; (3) information defined within the meaning of the provision of resources or objects; (4) production (individual or collective); (5) management and planning (activities and actors); (6) support and assistance; (7) emergence and systematization of meta-reflexive activity; (8) and “auto and hetero-evaluation” [PER 08, p. 20]. Each of these functions has a relationship with the others, and among all of them, it happens to be that some of them are perceived as primary by modelers compared to others. For example, the function of “information management” appeared central to Peraya and Bonfils [PER 14, p. 13] in a study conducted on PLEs (personal learning environments). In this study, “subjects tested instantiated five of the eight constituent functions of any publicized educational environment: it concerned these functions : (1) sharing ongoing issues and resources, (2) management information, (3). awareness, remote social presence, embodied by notifications, (4) production of written and multimedia documents and (5) communication and interaction. Each of these functions are associated with one or more specific mechanisms” [PER 14, p. 13]. These functions recognized and explained by the modelers are examples of this first modeling stage that we are presenting here. They take place in a concept map built for this purpose (see Figure 5.8).

11 Peraya et al. [PER 08, p. 20] cite successively: “Paquette [PAQ 93]; Collins et al. [COL 94]; [BAS 98]; de Vries [DE 01]; [HEN 01]; Peraya and Deschryver [PER 02, PER 03, PER 04, PER 05]; Gauthier [GAU 04]”.

Modeling a DLE Perceived as a Complex System

189

Figure 5.8. Modeling of group working environments [PER 14, p. 26]

5.1.2. Recursive processes We have seen that the recognition of the terms and legends inhabiting the conceptual map of a system and serving to name each of the perceived phenomenon traits to model is done by using successive iterations. The modeler (or project team12 responsible for modeling) iteratively completes this map by crossing, one by one, the functional levels of an archetype model (nine levels) that we described in section 5.1.2.4. “The argument is to be considered a sort of progressive complexification of systemic modeling” [PER 14, p. 58]. Adjustments will be made 12 When we discuss the “modeler”, we will keep in mind the fact that it can also be a team responsible for this modeling.

190

Modeling of Next Generation Digital Learning Environments

by successive iterations in order to minimize as much as possible the distance between a representation built by the modeler and the representation of users13. The architect who designs a project for a client always proposes several versions of the plans before producing a final version (the one accepted by the customer). This is also true when a business manager conducts the analysis of its production system. The first alarming piece of information concerning its assembly line is not sufficient to declare the system as failed. The latter will be subjected to further tests which, iteratively, will confirm a possible recurring malfunction. In our case, the iterative approach is similar to a process that consists of going back and forth between the construction of the model and the information gradually provided by the project actors and information and amassed by the modeler. The latter thus builds a system representation ever closer to reality. Better organized and more complete versions gradually appear. As the philosopher and logician Jean Ladrière14 once said in 1974: “It’s a teleology that is built. There is no teleology established in advance, there is a process of learning through which a tentative first approach manages to draw more and more precisely its own path. Internal process of auto-finalization”. In this regard, we have already noted that the approaches and latest generation modeling languages such as object approach or the corresponding UML are organized around basic principles of which the first is based on an iterative and incremental approach. 5.1.3. Representation of the processes and functional levels of the system In some cases, the modeler can see that the number of processors quickly increases. This should lead him to groups of processors in families or categories of processors. They can be classified, interrelated and prioritized in a “parents” and “children” tree diagram. This is a way of proceeding, starting from recognition of actions or complex of actions (processors), allowing the construction of more or less specific categories used to group these processors into parts, taking into account their properties or attributes. “We can then differentiate the system into many subsystems or levels, each level capable of being modeled by its network and interpreted relatively independently as soon as the interlevel coupling interrelationships have been carefully identified” [PER 14, p. 54].

13 In Lemire [LEM 08, pp. 105–106]: system of a constructed world and system of a represented world. 14 Quoted by Serge Diebolt in Le Petit Lexique des termes de la complexité of the “Réseau Intelligence de la complexité”, available (in French) at: http://www.intelligence-complexite .org/.

Modeling a DLE Perceived as a Complex System

191

Figure 5.9. A conceptual map produced by a French project team discussing the elements of Ancient Greek society [LEM 08, p. 94]

For example, if we can recognize an essential DLE function in distance tutoring, we can also admit that it corresponds to the combination of a number of functions: technological support, content expertise, methodological advice, animation, assessment [DEV 10, p. 3]. We have seen that graph theory makes a range of useful representations available to systemic modeling. Gilles Lemire, with whom we have worked concerning issues in system modeling since the early 2000s, offers the following conceptual map in his book Modélisation et construction des mondes de connaissances [Modeling and construction of knowledge worlds] 15 . The classification of components explains “institutions of ancient Greece (political, cultural and economic life); this remains an interesting illustration of this process for the development of a world of knowledge to construct” [LEM 08, p. 93].

15 The book for which we had the privilege of writing the foreword.

192

Modeling of Next Generation Digital Learning Environments

“In topological terms, the core of the conceptual map brings identification of the ‘world’ in which self-construction is undertaken; the first circle is fragmented in view of large classes, and the one that follows according to the number of subclasses. Specificity levels can stretch to terms representing concrete beings placed in the space that surrounds the last circle” [LEM 08, p. 93]. Other functional representations are possible. Two examples are shown in Figures 5.10 and 5.11, respectively.

Figure 5.10. Example 1 of a hierarchical representation

Figure 5.11. Example 2 of a hierarchical representation. For a color version of this figure, see www.iste.co.uk/trestini/learning.zip

Modeling a DLE Perceived as a Complex System

193

5.1.4. The eight constituent functions of a DLE applied to systemic modeling It is also possible to represent the general process of a DLE using, for example, the eight functions identified16 by Peraya [PER 03], Meunier and Peraya [MEU 04], Charlier et al. [CHA 06] as well as Peraya D., Charlier B., and Deschryver N., [PER 08, p. 20]. The exercise simply consists of completing the blank conceptual maps of the above representations. As Le Moigne recalls: “The General System is somehow a matrix whose modeler will establish, by molding a preconceived footprint (‘framing’ phase: general-modeling system) isomorphy. It then has a blank systemic model, without a legend. The ‘development’ phase will precisely involve writing captions, in other words establishing correspondences between the features of this systemic model and perceived or conceived traits of the phenomenon to model. The model ‘with the legend’ will thus be systemic (it is molded on the preconceived system)”17 [LEM 99, p. 41]. Figure 5.12 shows cases where these functions are applied to the diagram in Figure 5.8.

Figure 5.12. The eight constituent functions of any publicized educational environment [PER 08, p. 20] written at the heart of the conceptual map of the general process of a DLE 16 For the record, Le Moigne [LEM 99, p. 48] tells us that a process “can be represented by the designation of identifiable functions that it exercises or are can exercise”. 17 Le Moigne goes on to say that “its incompleteness [speaking of the model] will not be a regrettable imperfection, but a necessary condition for the anticipation, by simulation, of possible emergences of new behaviors within this complex system (interpretation phase 3)” [LEM 99, p. 41]. Here are some of forward-looking elements which will guide us.

194

Modeling of Next Generation Digital Learning Environments

But it would also be possible to integrate these eight functions (or a part of them) into any other graphic representation (Figure 5.13, for example), based on the models we have just proposed in Figures 5.10 or 5.11, for example.

Figure 5.13. An application of the constituent functions in previous models. For a color version of this figure, see www.iste.co.uk/trestini/learning.zip

We saw in sections 3.4 and 5.2.5 that there are complex system modeling languages that offer a range of symbolic representations, which are used to model these processes effectively. They are generally available from a library specific to the software used. Figures 5.14 and 5.15 give representations borrowed from UML (Unified Modeling Language), mainly resulting from “object-oriented” modeling (OMT) by Rumbaugh et al. [RUM 95]. We have also justified this choice in section 5.2.5. In this example, we recognize the eight processors18 referred to earlier in Figure 5.12: – awareness; – support and tutoring; – management and planning, etc.

18 UML packages.

Figure 5.14. Eight processors in UML (or packages). For a color version of this figure, see www.iste.co.uk/trestini/learning.zip

Modeling a DLE Perceived as a Complex System 195

Figure 5.15. An extract from the support and tutoring use case built for Project Management MOOCs (see full diagram in section 6.1.1). For a color version of this figure, see www.iste.co.uk/trestini/learning.zip

196 Modeling of Next Generation Digital Learning Environments

Modeling a DLE Perceived as a Complex System

197

NOTE.– Here, we will only provide incomplete views to illustrate our point. They will be fully developed in Chapter 6. The box surrounding these processors corresponds to the black box (in blue), or the encompassing processor, in other words, the system (Figure 5.14). Figure 5.15 represents the “children” processors of the supporting and tutoring processor as an example. In UML language, the diagram in Figure 5.14 represents the functional decomposition of the system processors also called “packages” in this language (another meaning for the same word!)19. This diagram can also indicate the actors involved in each of the packages. The diagram in Figure 5.15 represents children processors (also known as “use case” in this language), functions more specifically necessary to users. In UML, it is possible to make a use case for a DLE chart as a whole or for each package. Here, the use case bears the “support and tutoring” processor. 5.2. Synchronic processes (the system performs) Processors registered in the heart of the concept map20 will now be operated in turn, taking into account the actions they produce (explaining how the input values (inputs)21 of each processor are transformed to become output values (outputs)). To model an active environment involves modeling the activity exercised by the fulfillment of actions, transactions and interactions. This step corresponds to the second level of archetype model (nine levels) complexification. “This identifiable phenomenon is perceived actively: it is perceived because it is presumed to ‘do’ something. In terms of modeling techniques, we go from the ‘ticker’ symbolizing a closed set to the ‘black box’ symbolizing an active processor” [LEM 99, p. 59]. We immediately add a level of additional complexity that the designer is able to perceive and represent simultaneously: the self-regulation of the identified processes. This is located on the third level of archetype model complexification. 19 In fact, Le Moigne is already using process terms, functions, actions, complex of actions and processers to mean the same thing. 20 For which the envelope is precisely the “ticker” mentioned above. 21 Note that a data flow diagram of the functional model of object-oriented modeling that we use for the situation “shows how the output values are derived from input values” [RUM 95, p. 124].

198

Modeling of Next Generation Digital Learning Environments

“To be effectively identifiable, the phenomenon must be perceived by some form of consistency or stability. Its behavior, because it is perceived, is presumed to be ‘regulated’: we don’t model the chaos or the erraticism! In other words, the modeler assumes the emergence of some internal regulation mechanisms” [LEM 99, p. 59]. To highlight the presence of actions and counter-reactions, in other words, interrelationships of actions (network and feedback), we suggest indicating their active presence (section 5.2.1) by recording their origin and their destination in a structural matrix. To translate the presence of regulation within a system at a given time (the system that performs and seems regulated), we will draw up the inventory of input and output values for each of them (section 5.2.2). Then, to “account for the growing complexification of this regulation” [LEM 99, p. 60], placed at the third level of archetype model complexification, we propose the construction of a data flow diagram (section 5.2.3) with the ability to highlight the information produced by the system, but also its intermediary forms and the systems of symbols, providing intermediation and regulation, similar to the way “of presumed electric impulse that through nerves or a circuit, transmit regulation instructions, reflecting changes in the system’s state” [LEM 99, p. 60]. We will also show that the digital learning environment that we model is also able to inquire about its own behavior. In the section that follows, we will address fourth level of archetype model complexification, not only statically (the system “performs”) but according to its dynamic component (the system “becomes”). We will describe the typical phases of the processes (scenarios), highlighting the events that activate the actions (section 5.3.1). The traces left by these events will be helpful in developing the state diagrams of the processes involved. We will finally offer a description of each of these functions (processor) (section 5.3.4) before specifically focusing on a DLE’s ability to process information, decide its own behavior and remember information (see section 5.4). This progressive complexification of systemic modeling corresponds, respectively, to the fifth and sixth levels of an archetype model. 5.2.1. Identifying the active presence of interrelationships Let us start by outlining a simple method allowing the identification of the active presence of interrelationships which includes feedback, also known as “recycling”.

Modeling a DLE Perceived as a Complex System

199

The structural matrix built for a similar purpose by Le Moigne [LEM 99, p. 50] is resumed at good cost in the modeling approach that we propose. “The complexification of the system modeled will be made by an interrelation of the previously identified N processors by the composed functions that each provides. It is said that there is interrelation between two Pi and Pj processors when an output (or an output vector component) Pi processor is an input to the Pj processor. IR interrelation (Pi, Pj) is then enabled”. “All combinations of possible interrelationships among N processors can be represented using the structural matrix of the system (representation that can be refined by the creation of the N2 matrices from PiPj coupling in the case where the Pi outputs can be the inputs of several different Pj, Pk, Pl processors...)”.

Figure 5.16. A matrix of possible interrelationships between N processors [LEM 99, p. 50]

“The presence of a 1 in the O/P Pi, I/P Pj box means that the Pi, Pj interrelation is enabled; the presence of a zero means that this interrelationship is not enabled, and possibly that it is prohibited. This Boolean structural matrix allows the economic presentation of several interesting characteristics of a complex system” [LEM 99, p. 50]. In addition to the interest that this simple representation contains when viewing the system interrelations formally and quickly, it also has the advantage of benefiting from its integrating effect to “show new behaviors, rarely predictable by

200

Modeling of Next Generation Digital Learning Environments

linear composition 22 ” [LEM 99, p. 50], which we have seen in section 5.2.6, allowing a connection with “our” prospective projects put into perspective. But it is true that in practice, modeling a complex system is rarely done by starting with the construction of the structural matrix because the preconceived identification of N processors and input-output pairs by which they are characterized... is not given by the problem that the modeler considers [LEM 99, p. 54]. Therefore, it will rather be starting from known system projects that SM will propose to initiate in the model’s design process. 5.2.2. Establishing an input and output value inventory for each process From a methodological point of view, and in order to facilitate the construction of an input and output inventory and the recognition of functional dependencies that follow, we propose adopting a method inspired by “object-oriented” (OMT) 23 modeling developed by Rumbaugh et al. [RUM 95]. In this particular IT modeling paradigm, an object is nothing other than a simple constituent of the system. This clarification is important because it keeps us safe from a situation that could become sensitive to SM choice (vs. AM), a choice extensively explained in the preceding chapters. With this IT paradigm, the object represents a concept, idea or any entity from the physical world and not only a static grantor or a body of the system. And if the object can represent a concept, a process can be an object in this sense since it is itself a concept 24 . Moreover, object-oriented methods also address the issue of modeling complex phenomena perceived as complex by their dynamic and functional aspects, derived from modeling models of the same name to which we will have the opportunity to revisit. In an object-oriented approach and in order to identify the input and output values, one usually represents the process by “a flowchart” (or a data flow diagram) which allows the output values from input values to be obtained. For a determined process state, it is a matter of specifying its activity and determining the organization of operations that define it, explaining data flow that circulate during these operations.

22 Hence, their name of counter-intuitive effects. 23 Object Modeling Technique. 24 Sfard, 1992, quoted by [MIN 12, p. 3], calls “‘process reification’ the transition from ‘process’ design to ‘object’ design. The author also pointed out that the process and object terms should be understood as different facets of the same notion rather than as separate components. Both operational and structural aspects are therefore complementary”.

Modeling a DLE Perceived as a Complex System

201

Figure 5.17. The identification of input and output values for a given system state

The identification of input and output values occurs at the time of a fine analysis of the constituent action sequences of any process. It is a task that consists of identifying the list of information carrier events that are then transformed into data transmitted to the system or issued by the system. This information transports input values, such as, for example, the identification of a user wishing to connect to his DLE (an MOOC, for example). This alphanumeric information becomes an input value that tells the system that the user is registered (or not) to the course and that the latter seeks approval to integrate the activities to which he is registered. Screens of different types or input (or output) windows (input mask) of information can result from this approach and take place on the screen depending on the processing that the system will perform from this information. Here, the general activity is learning, and the goal is to enter the system. The system interface is to allow the user entry by referring relevant information based on the information received. For example, the system can accept to host the user if the data provided is consistent with the data previously established by a “decision-making system”, which will itself have provided this data to a “memory system” providing a storage function of information. Input values may be the identifier and the password, for example, and the output values the message “password or identification invalid: try again” or even “a new web interface home page”. 5.2.3. Representing data flow and functional dependencies between processes On the fourth level of archetype model complexification, “everything happens as if the system has endogenously produced intermediary forms, information, symbol systems, which would provide regulation intermediation” [LEM 99 p. 60]. To

202

Modeling of Next Generation Digital Learning Environments

account for the growing complexity of this regulation, it seems reasonable to propose sketching a data flow diagram. From the approach of object-oriented modeling (OMT), this diagram graphically represents the flow of data between the various processes in the system. “It allows the determination of its borders, the area of study, the system processes, processing (activities) and gives a concrete view of the system to build”25 or analyze. It is also used to show functional dependencies. It should also be noted that this diagram “is interested in processing data without regard to sequencing ‘decisions’ or the structure of objects” [RUM 95, p. 178]. Its interest is to show how the output values are obtained from the input values, or how these values are processed, and how the system will behave. We have said that this is also the moment when new behaviors within the system can emerge because of this connecting of processors.

Figure 5.18. The highest level flow diagram of an MOOC interface

25 Definition taken from a presentation (TM070-v110a-2015-03-08), dated March 8, 2015, from Luc Lavoie of the Faculty of Science of the University of Sherbrooke, available (in French) at: http://info.usherbrooke.ca/llavoie/enseignement/Modules/TM070-DFD.pdf.

Modeling a DLE Perceived as a Complex System

203

In practice, this diagram is usually built in layers which successively refine non-trivial processing: any non-trivial processing must be described by a sub-diagram. “The highest level layer may consist of a single treatment or may be a treatment to collect entries, another to deal with data, and one to produce the outputs” [RUM 95, p. 180]. Figure 5.18 shows the highest level data flow diagram of a MOOC interface, taken as an example, and considered here as a latest generation DLE (the legend is in Appendix 1). Figure 5.19 develops the process, producing the outputs of Figure 5.18 and gives the example of a non-trivial data flow sub-diagram of the general diagram. In the same way, this sub-diagram can be developed in other sub-diagrams. The processor showing the forum will certainly give rise to the development of another data flow diagram and so on. In diagrams, “features are identified by the verbs used to account for the events to update. They become the opportunity for a relationship of association between entities: each time, this creates a ‘functional dependency’” [LEM 08, p. 151]. Relationship cardinality characterizes the relationship between an entity and the relationship. Functional dependency is said to be strong for a cardinality (1,1) and weak for cardinality (0,1) [LEM 08, p. 152]. “Cardinalities allow for the characterization of links that exist between an entity and the relationship to which it is connected. Relationship cardinality consists of a couple with a minimal and maximum point, an interval in which the cardinality of an entity can take its value: (i) the minimum point (usually 0 or 1) describes the minimum number that entity can participate in a relationship; (ii) the maximum point (usually 1 or n) describes the maximum number of times that an entity can participate in a relationship. A 1.n cardinality means that each entity belonging to an entity class participates at least once in the relationship. A 0.n cardinality means that each entity belonging to an entity class does not necessarily participate in the relationship” [LEM 08, p. 152].

Figure 5.19. A processor data flow diagram "producing” MOOC OpenClassrooms “outputs”

204 Modeling of Next Generation Digital Learning Environments

Modeling a DLE Perceived as a Complex System

205

Also note that the represented DLE equally contains objects used as information storage (Figure 5.18). This is the case here in the registered database that most notably stores their usernames and their passwords. “Data reservoirs differ from data flow or processing by the fact that the entry of a value does not result in an immediate exit, but rather this value is reserved for future use”. This memory system will be discussed on the sixth level of archetype model complexification Furthermore, although the data flow diagram (Figure 5.20) shows “all the possible ways of processing” [LEM 08, p. 181], they are usually waived decisions. Like the memory system, the decision-making system will be studied in detail later and especially when we get to the fifth level of archetype model complexification. Decision functions, such as, for example, those who verify a username or a password, obviously influence the result. Some data values can exclude certain intended actions or affect the outcome of a decision. “Decisions do not directly affect the data flow model since this model shows all the possible ways of processing. It may however be useful to enter the decision functions in the data flow model to the extent that they can be complex functions of the input values. Decision functions can be represented in a data flow diagram, but they are indicated by dotted outgoing arrows. These functions are only indications on the data flow diagram; their result only affects the flow control and not the values themselves” [RUM 95, p. 181]. Here is an example:

Figure 5.20. A data flow diagram integrating decision functions

More generally, in an object-oriented (OMT) approach, this set of data flow diagrams was built in order to clarify the meaning of operations and constraints,

206

Modeling of Next Generation Digital Learning Environments

constituting what was agreed to be called the functional model. It is interested in data processing without regard to “sequencing, nor decisions, nor structure of the objects” [RUM 95, p. 178]. It shows how “output values in a calculation are derived from incoming values regardless of the order in which they are calculated” [RUM 95, p. 124]. We will see in Chapter 6 that the current object-oriented modeling languages like UML (which are actually just an extension of OMT) reflect this same intent with the so-called “use case” and “activities” diagrams. Use case diagrams provide an overview of a system’s functional behavior (corresponding to the OMT functional diagram); these activities allow a focus on processing. They are particularly suitable for modeling the data flow and allow us to graphically represent the behavior of an action or the progress of a use case. We will mainly use these diagrams (derived from UML) in our modeling projects. We have provided arguments in section 5.2.5. They characterize a system “that does” clearly indicating the actions. Remember that the only reason leading us to previously introduce diagrams specific to OMT “techniques” or “methods” before those of UML stems from the fact that we recognized a founding language in OMT (possessing wanted assets), which developed current languages like UML in its different versions, UML 2 and SysML, etc. 5.3. Diachronic processes (the system becomes) We have just represented the development of data registered at the heart of a finalized process circulating in an active environment. The data flow diagram shows functional dependencies between these values. At this third level of complexification, this helped highlight interrelationships among processes at a precise moment in time; in a static way: “the system performs”. But the canonical process model [RUM 95, p. 47] shows that it is also necessary to represent the system that transforms by “acting”, over time (see Figure 5.21).

Figure 5.21. The canonical process model [LEM 99, p. 47]

Modeling a DLE Perceived as a Complex System

207

To arrive here, it is a matter of reaching the fourth level of archetype model complexification. In Rumbaugh et al.’s object and object-oriented modeling [RUM 95, p. 86], “these system aspects, dependent on time and changes, are grouped in the dynamic model...”. If we compare this model to the previous, we could say that if “the functional model indicates what is happening, the dynamic model indicates when this happens...” [RUM 95, p. 124]. We therefore propose that the modeler adopt this principle to explain, in a synchronic way, this “transforming” action in time. It is thus encouraged to describe the typical process phases (scenarios), highlighting the events that activate the actions. The traces left by these events will be useful in developing state diagrams that specifically account for expected “system state changes” at this level of complexification [LEM 99, p. 60]. Remember that at the fourth level of archetype model complexification: “everything happens as if the system has endogenously produced intermediary forms, information, symbol systems, which would provide regulation intermediation [see RUM 95, p. 109] (thus the alleged electric impulse passing through nerves or a circuit transmits ‘regulation instructions’, ‘reflecting changes in the system state’). This symbolic emergence of information, artifact or internal communication device, constitutes an ‘event’ or jump, in the presumed complexification of the modeled system” [LEM 99, p. 60]. As we noted in the introduction to section 5.2, we approach this fourth level of complexification no longer in a static manner (the system “performs”) but according to its dynamic component (the system “becomes”). Notice again that the data flow diagrams previously proposed and constituting the DLE functional model respond to the first requirements at the fourth level of complexification. “The endogenous production of intermediary forms, information and symbol systems that provide regulation intermediation” seems well represented in these diagrams (information, pictograms, process input and output). The DLE dynamic model that we will now develop will be able to justify changes in the system state at this level of complexification. We will begin by describing the typical process phases (scenarios), highlighting the events that activate the actions and transform them. We will then look at the traces left by these events. The latter will be useful for developing state diagrams at the end of section 5.3.3.

208

Modeling of Next Generation Digital Learning Environments

In object-oriented (OMT) modeling, diagrams represent the traces of action control which manifest through the translation of information exchanged and transformed during these activities. 5.3.1. Typical sequences (or scenarios) and even tracking The word “scenario” has many meanings that adjust to the context in which it is used: cinema, management, education and informatics. Linking it to the context that we are concerned with (education/informatics), a scenario is described as a “set of events or behavior to describe the possible ‘evolutions’ of a system” 26 . This definition is not very far from the sense that books give on object-oriented modeling techniques: “a scenario is a sequence of event types; it allows for the description of common interactions for the extraction of events and the identification of target objects” 27 . The diagram is again the preferred instrument to represent the interface components and the system, as well as the courses that the users of the mechanism follow. “Each custom diagram becomes a bearer of cognitive activities that will place the user and the system in favorable contexts to interaction realization” [LEM 08, p. 132]. This leads us to specify actions performed on objects: this is the case, for example, when we click on an icon to specify an event type and the kind of cognitive activity that one maintains with the system. This can be the display of an illustration offered by way of explanation, that of a graphic text, or opening a window, allowing writing in order to communicate an opinion. The idea would be to make as many diagrams as there are events that will put the user in the presence of information or communication, either with the system or with others [LEM 08, p. 132]. Here, we will not give an example of diagram scenarios specific to object-oriented (OMT) modeling because we will have the opportunity to do so later when a specialized modeling language has been adopted (UML 2.0). We will then see that, in this language, the scenarios themselves are specific diagrams inside “use case” diagrams or “state transitions”. With that said, Lemire [LEM 08, pp. 132–136] proposes some examples of schematic scenario representations and a number of tips to build them. The reader can easily refer to his work if interested. 5.3.2. Traces left by events The methodology implemented is based on the traces left by each of the activities, whether they are cognitive, informational, decision-making, 26 Grand Dictionnaire terminologique. 27 Alissali M., Introduction au Génie Logiciel (course material), 1998, available (in French) at: https://fr.scribd.com/document/10176354/GenieLogiciel.

Modeling a DLE Perceived as a Complex System

209

organizational or regulatory. They are retrieved from surveys, semi-leading interviews with various actors of the considered DLE. We have collected much data on MOOCs upon their arrival in France in order to model these specific digital learning environments [TRE 15a, CHE 15a, CHE 15b]. This data tells us both about the perceptions and behavior of actors but also those of the registered users (students). This body of data can also result from specific computer applications introduced in DLEs in order to monitor training and its various actors (potential dropouts) online. These computer applications often provide accessible data in real time via a dashboard developed for this purpose. The BoardZ application that we will see later (Figures 5.28 and 5.29) is a good illustration of this. We will return in detail to this application in section 6.5.2 in Chapter 6 when dealing with data provenance and collection. To collect these traces, we suggest combining two different approaches. The more pragmatic approach consists of collecting digital traces from learners with tracking tools. Often customizable, these tools allow for the modeling of the type of learning that you want to monitor. The goal is to be able to follow an area of study, be it from the teacher or student’s point of view, or even the institution’s. The data is often presented in the form of diagrams in educational monitoring “dashboards”28. The other approach is more traditional as it is based on surveys or semi-leading interviews. 5.3.3. The states and state diagrams We resort to this to form the concept of state diagrams specific to Rumbaugh et al.’s “object-oriented modeling” 29 language [RUM 95, p. 91]. These diagrams allow for the representation of action control traces which manifest through the translation of information exchanged and transformed during these activities30.

28 For example, the “BoardZ” project under development and testing, available at the following address: http://moodlemoot2014.univ-paris3.fr/course/view.php?id=211. 29 James Rumbaugh is the creator of the “object-oriented modeling” language”. He is also one of the three fathers of UML with Ivar Jacobson and Grady Booch (source: online encyclopedia Wikipedia). 30 Note: the processing regarding a data flow diagram corresponds to the activities or actions on state diagram classes. The flow, in a data flow diagram, corresponds to the objects or trait values on a state diagram.

210

Modeling of Next Generation Digital Learning Environments

Figure 5.22. A state diagram constructed from an OpenClassrooms MOOC

In the same way that we stated for functional OMT diagrams in section 5.2, we can repeat this for OMT state diagrams above: the current object-oriented modeling languages in UML are able to translate this intention. In this precise case, they are concretely “state transition” diagrams that we have used in our “modeling projects” (see section 7.1.3 in Chapter 7 and Figure 6.12 in Chapter 6). We have given the arguments in section 4.2.5 in Chapter 4. They characterize an “evolving system” clearly indicating the different processor states. 5.3.4. Process descriptions Remember that any complex system can be represented by a system of multiple actions or by a process that can be a tangle of processes, each of which can be represented by the designation of identified functions that it exercises or may exercise. The previous steps have allowed the designation of roles that these processes take, thus revealing values circulating in the mechanism to build and that are also exploited.

Modeling a DLE Perceived as a Complex System

211

To portray them, the modeler focuses on what the system does and seeks to give an intelligible representation. This can be done in different ways: in natural language, pseudocode, mathematical formulation, using decision tables, etc. Rumbaugh [RUM 95, p. 181] recommends that the modeler “focus on what the function should do, not how to implement it. [...] The main goal is to show what happens in all cases”; thus, in the event of a free registration of an Internet user to an OpenClassrooms MOOC, the learning pace imposed (over 4 weeks). 5.4. A system capable of processing information and deciding its own behavior A fifth level will now be added to the steps of systemic modeling complexification. The system becomes capable of deciding its own activity, processing the information that it produces and making decisions on its own behavior. “Capable of generating information, the system becomes more complex by proving itself capable of handling: symbol computation, then cognitive exercise, the system becomes capable of developing its own behavioral decisions. It is necessary now to recognize an autonomous decision subsystem, processing information and only information in the model of the system” [RUM 95, p. 60]. The fifth level is a milestone in the progressive nature of the nine-level archetype model complexification process. The first four levels characterize cybernetic and structuralist procedures: the system has a project in an active environment; it exists, performs and transforms. The following levels first show the system’s ability to generate, process and store information (levels 5 and 6). Then, they identify it as capable of coordinating (level 7) and developing new projects and new action forms and being imaginative (level 8). Finall0079, the active process develops an auto-finalization ability which allows it to decide its own future and to make choices regarding its own direction (level 9). 5.5. A model based on analysis data 5.5.1. A model constructed from analysis data The model’s construction is performed by successive iterations between the system of a constructed world and what the users perceive (their viewpoints) when they observe the realities or objects that constitute this world31. 31 Having only a perception of the real world, we cannot take it as a reference.

212

Modeling of Next Generation Digital Learning Environments

“Possible discrepancies between the ‘system of the constructed world’ and the ‘system of the represented world’ should be identified at the time of audit from users of lesser or greater compliance (systems of the constructed and represented world). If adjustments are needed, they must be by successive iterations in order to minimize the distance between these two worlds as much as possible” [LEM 08, p. 108]. “Our vision of the world is a model. Any mental image is a model, fuzzy and incomplete, but serving as a base for decisions” [DER 75, p. 122]. From a first draft of solutions, systemic analysis proceeds by a permanent back and forth between a model to be built and the analysis data results provided by the system. “It is in this back and forth between data acquisition on the basis of modeling hypotheses and their reconstruction by modeling that a science of complex systems modeling can develop” [BOU 04b, p. 1]. The raw data collected is analyzed beforehand (data analysis) by applying such statistical procedures having recourse to multivariate techniques32 in view of system complexity and data variety. These techniques allow for the return of more synthetic and therefore simpler results by identifying and grouping the variables or terms that are similar (groups of “correlated” variables). These results constitute the analysis data which precisely allows for model refinement (examples to follow). The model is gradually constructed, taking into account this analysis data that reduces the distance between the systems of the constructed and represented world at each iteration. “‘Modeling’ consists of building a model from the systems analysis data” [DER 75, p. 122]. 5.5.2. Provenance and data collection The data to analyze can be of all kinds: cognitive, informational, decision-making, organizational or regulatory. It can be surveys or interviews with different actors of a DLE or computer systems that can restore traces from a learner activity, like, for example, dashboards located on course platforms or tracking tools that allow for the identification of registered user profiles. It can also include simulation results that the modeler has decided to use in order to refine his model. All this data, produced and gathered by the modeler, allows for the determination of a succession of approximate solutions (analysis data) that are gradually approaching the perceived reality. It is therefore up to the modeler to determine what data is to be collected and retained, data that seems best adapted to the analysis context. 32 Factorial principal component analysis (PCA), factorial multiple correspondence analysis (MCA), ascending hierarchical classification, etc.

Modeling a DLE Perceived as a Complex System

213

According to Cisel, for example, “Early research on MOOCs is to a large extent based on learning trace analysis that can accurately measure what is happening on the teaching platform. But a purely descriptive approach of these usages is a problem if you want to understand the ins and outs of these behaviors on one side, and what is happening outside on the teaching platform on the other. To address these issues, more qualitative approaches derived from human and social sciences are essential: surveys, interviews, field observations; would it not be important to answer this simple question ‘Why subscribe to an MOOC?’” [CIS 16] If the data provided by the system is plentiful and varied, the most relevant data to analyze is usually suggested to the modeler by the model itself. For example, the existence of a “forum” function registered at the heart of a model may encourage the modeler to want to know the reasons that motivate viewers to associate with it. He can question people affected by this mechanism in the form of interviews or questionnaires. These various kinds of data may possibly allow the modeler to discover that the forum function originally scheduled by the designer was transformed or converted, or simply that it has not been used... Conversely, the data analysis stemming from the system may reveal defects in the model’s internal structure which would correspond to the analysis data. This should alleviate the problem. There are several non-exclusive data collection methods highlighted below and we will explain them in detail in the following sections. But if surveys, interviews and field observations are often preferred in the human and social sciences field, other means of collecting this data (tracking, simulation, etc.) can be useful to system analysis and to model improvement. Among these analysis methods, the following are included: – analysis from data provided by tracking tools; – analysis from data provided by dashboards; – analysis from data provided by simulation; – analysis from data provided by questionnaires and interviews. We will develop each of these methods in the following sections, and in Chapter 6 we will see concrete cases of modeling and analysis.

214

Modeling of Next Generation Digital Learning Environments

5.5.2.1. Data provided by tracking tools Establishing activity traces involves highlighting the flow of information over communication activities, to show where the information is coming from and where it is going, the knowledge of the places where it is acquired, to specify how it is produced and operated, and to verify if the information flow is as planned and if communication is done according to expectations [LEM 08, p. 140]. There are several tracking tools which allow the participants’ actions in a DLE to be followed. The latter being specifically built on a web interface makes it easy, upon each visitor’s action, to send information using a tracker (a script 33 ) to a statistical program that is installed on the web server and to store another part on the visitor’s computer via cookies (browsing history, identification information, etc.).

Figure 5.23. A web tracking circuit

To perform this tracking, there are still several tools. The best known and most commonly used is Google Analytics. It writes the tracking code “in the websites’ source code in the location recommended by the statistical tool and on all pages of the site. If the website is made with a CMS, it will be necessary to insert this code in the page templates or functions called by the theme” 34 . The information is then processed by the statistical tool, and the webmaster can then view the results. 33 A program or a piece of software that will run a function at the time of a web page display or the realization of a user action. 34 Excerpt from the Dictionnaire du Web, available (in French) at: https://www.1min30.com/ dictionnaire-du-web/tracking.

Modeling a DLE Perceived as a Complex System

215

For example, here is what a tracking code in Google Analytics looks like:

Figure 5.24. A tracking code in Google Analytics

With MOOCs being perfect representatives of the latest generation DLEs, they generate, just like other learning environments, traces of online activities that one can retrieve and process. In the second edition of MOOC EFAN (Enseigner et former avec le numérique), Boelaert and Khaneboubi [BOE 15] “counted more than 4,800 registered users and almost 500 activity deposits. The participants’ actions online generated traces of connections on MOOC servers (web logs) [...]. The logs are non-declarative information concerning those who log on to a website”. For them, it concerned “potentially interesting data in order to identify, for example, learner trajectories, typical behaviors or simply executing audience measurements” [BOE 15]. “[The recovered logs enabled them to] define five states for each userweek couple: –‘absent’: the user is never connected to the resources corresponding to this week; –‘present’: the user has connected to the pages of the week, but has not watched videos nor answered quizzes (he could download videos, or read, or simply visit pages to follow links to other sites); –‘video’: the user has watched at least one video of the week, but has not answered the quiz (unfortunately, the edX logs do not know if the user has uploaded a video, we know only if he looked at it directly on the course site); -‘view quiz’: the user has consulted the quiz of the week, but not answered it;

216

Modeling of Next Generation Digital Learning Environments

-‘quiz’: the user has answered at least one question from the quiz of the week” [BOE 15]. Considering these representative states of degrees of investment in the MOOC, they draw “chronograms” for each week of courses and conclusions on these results. Figure 5.25 represents some of the participants who were in the five previous states, and Figure 5.26, constructed from the same data, represents the individual trajectories of all registered users in the form of a “carpet”35. The trajectories are arranged according to their end trajectory state [BOE 15].

Figure 5.25. Frequency of the different states, week by week [BOE 15]. For a color version of this figure, see www.iste.co.uk/trestini/learning.zip

35 “This visualization reads as follows: each row is a user, that is followed from left to right starting at week S0 up to week S5. This visualization reads as follows: each row is a user, that is followed from left to right starting at week S0 up to week S5. On this graph, we have arranged the individuals according to their state in week S5; thus, the trajectories represented at the top of the chart are those of the participants who answered the quiz during the last week” [BOE 15].

Modeling a DLE Perceived as a Complex System

217

Figure 5.26. “Carpet” of the individual participants’ trajectories [BOE 15]. For a color version of this figure, see www.iste.co.uk/trestini/learning.zip

5.5.2.2. Data from reports generated by distance education platforms Another simple and effective method is the analysis of connection data from reports generated by DLE platforms. Analysis of the traces left on the platform allows the teacher to learn more about his students’ activity via a PHP program. Figure 5.27 gives an example of analyses conducted from 2013 to 2015 by Éric Christoffel, a Professor and research colleague in our laboratory (LISEC), who wanted to know and precisely study his students’ activity in L3 MPC (Math-Physics-Chemistry). He also produced four graphs that allowed him to draw some interesting results: – video playback in minutes per day; – number of clicks on the course’s video page per day (correlated with the previous graph); – number of course notes consulted per day (PDF course notes); – number of downloads of various resources (in PDF format): subjects and correction, etc.

218

Modeling of Next Generation Digital Learning Environments

Figure 5.27. Data collected from reports generated by Moodle [THO 08]. For a color version of this figure, see www.iste.co.uk/trestini/learning.zip

In the first three graphs, it can be noted that the students are active the night before an examination or when a flipped classroom is imposed. In the last graph, it shows that students are regularly active throughout the year when it comes to downloading other resources: TD, TP text, subjects of previous years and corrected. 5.5.2.3. Data generated from dashboards: BoardZ project as an example The Faculty of Avignon’s BoardZ project is an example of a generic tool located at the interface between the data collection mechanisms seen in the previous section and simulation. It is above all an educational dashboard engaged in distance learning. It is fully customizable with everyone possessing the ability to model the type of learning that he wants to oversee. One of the objectives of this project is to track the learning process of learners, in real time, in a distance learning environment. And it is from the teacher’s point of view, but also the student’s or the

Modeling a DLE Perceived as a Complex System

219

institution’s, offering an analysis of the digital footprints left by the learner. The project must therefore allow the teacher to make sure that everything goes “well” in his course, alert him as soon as possible in case of student dropout, give him a general trend of student practices, but also make an individual state by a student. In the same way, it gives the student the opportunity to position himself against the group’s practices and the teacher’s expectations. It must also allow the institution to “monitor” learning by presenting an overview of the practices in each lesson. More specifically, particularly for FC, it is possible to model a generic response to accredited fund-collecting agencies concerning an intern’s course follow-up. The challenge is to be able to extract digital traces regarding learners’ activity to prove to accredited fund-collecting agencies that they have actually followed the course. In summary, the educational dashboard project is a multitool allowing for: – the detection of early student dropouts; – the verification of consistency between the learners’ practices and the teacher’s expectations; – the facilitation of an individual learner monitoring in a course taught on-site or remotely; – the provision of evidence given by accredited fund-collecting agencies in lifelong education. “BoardZ is an open source tool developed in Avignon that allows for the geographical viewing of educational indicators within a Moodle dashboard”. Several uses of this dashboard are possible depending on the user’s profile. Each profile has a progressive approach: from global to detailed. The global view allows for the comprehension of the whole area at a glance, while the following two levels offer a finer modeling of the objects of study selected. The viewing platform is generic and allows everyone to (re)define their own indicators by simply (re)writing SQL queries or redefining the functions to obtain the indicator values to customize. The range that we present takes this data based on Moodle, but it can look for data in the school’s SI. For example, the teacher profile offers two previews for each of its courses on the first level, modeling ICT uses for one and the learners’ activity for the other. The second level presents the selected course activity of each learner in relation to others. Finally, the third level of precision displays indicators specific to one student for this course.

220

Modeling of Next Generation Digital Learning Environments

For each level, a warning system is available allowing the teacher, for example, from the first level of the teacher profile, to know that a learner has dropped one of his classes. In addition to the teacher profile there are: “student, training manager, platform administrator, ICT jury labelling profiles” [MAR 15]36. According to Fanny Marcel and Thierry Spriet who are behind this project, this dashboard aims to promote student success by identifying dropouts as early as possible. They start from the realization that the Moodle platform reports are rather oriented toward individual monitoring, far from mass teaching. This dashboard must display an individualized profile by presenting relevant indicators while highlighting the abnormal behaviors. Pedagogical modeling combines several criteria (number of connections per week, access to resources (files, pages, etc.), activity on the forum, wiki, database, homework records, tests, etc., use of the type of graphic representations as seen in Figure 5.28).

Figure 5.28. A pedagogical BoardZ modeling combining five criteria

Access to different dashboards is possible: students/teachers/course managers. Each program actor is associated with a custom modeling. For example, the teacher will find a modeling type of each of its courses, activities, student dropout alerts, an overview of student activity and the ability to track each student individually. Figure 5.29 gives a brief overview of a French course list with a title, a description of the alerts and lists of students and models. 36 http://moodlemoot2014.univ-paris3.fr/course/view.php?id=211 ([email protected]).

Modeling a DLE Perceived as a Complex System

221

Figure 5.29. A project BoardZ pedagogical modeling. For a color version of this figure, see www.iste.co.uk/trestini/learning.zip

The goal today is to refine these pedagogical models, to integrate evaluations into their calculations and allow configuration of the course models (with the teachers’ expectations, etc.). Work groups in which we have participated are currently hard at work to complete this project. 5.5.2.4. Simulation data According to Rosnay [DER 75, p. 122], “simulation ‘tries to provide’ for a system allowing the simultaneous game of all variables; what limitations of our brain prohibit without computer assistance or simulation devices. Simulation is based on a model, itself established from prior analysis”. “System analysis, modeling and simulation are thus the three fundamental steps in the study of the dynamic behavior of complex systems” [DER 75, p. 122].

222

Modeling of Next Generation Digital Learning Environments

Simulation is interesting when observing complex phenomena, because it allows for the simultaneous variation of variable groups, just as happens in reality. It is then possible to speculate hypotheses on observed phenomena, their behavior, to test initial conditions and to study the responses and reactions by varying them. Simulation is therefore able to provide real-time responses to the various user decisions and actions; however, it requires powerful computers and, in some cases, cumbersome experimental editing. For example, to learn more about the influence of the wind on the lives of trees and plants, Emmanuel de Langre, Professor of mechanics at the École Polytechnique, and Pascal Hémon, a research engineer and aerodynamicist, installed a tree in a wind tunnel for the first time!37 From a DLE point of view, simulation does not require cumbersome mechanical means, but nevertheless, it requires the use of powerful computers capable of processing real-time variations of the multiple variables at play. Simulating a complex system goes through a preliminary step of modeling system constituents, from their behaviors and interactions between these constituents and with their environment. The steps of this model structure have been described in detail at the beginning of Chapter 5. The next step consists of running the obtained model by mobilizing the appropriate computer means. “One of the characteristics of these systems is that one cannot anticipate the evolution of the system modeled without going through this phase of simulation. The ‘experimental’ approach by simulation allows for the reproduction and observation of complex phenomena (biological or social, for example) in order to understand and anticipate their evolution”38. Implementation is possible through behavioral simulations and multi-agents. Applicable to social sciences (social networks simulation), they allow for, in particular, the illustration of a number of perspectives from this domain and the interest of multidisciplinary approaches to address such systems. There are many tools for the simulation of complex systems such as the multi-agent simulation platform (Netlogo, RePast, Gama, etc.), the dynamic systems platform (Stella) or even Anylogics, etc. As an example, Figure 5.30 illustrates an IRIT39 simulation in the multi-agent Netlogo simulation interface40. 37 https://www.polytechnique.edu/en/modeling-and-optimization-of-complex-systems. 38 An excerpt from a Master 2 course, Research–Computing and telecommunications, under the guidance of Frédéric Amblard of the Institut de recherché en informatique de Toulouse (IRIT) of the University of Toulouse III-Paul Sabatier, available at: https :// www.irit.fr/ M2RIT/. 39 Institut de recherché en informatique de Toulouse.

Modeling a DLE Perceived as a Complex System

223

5.5.2.5. Data provided by surveys, interviews and field observations We have just seen that research based on the analysis of trace activities or simulation results allows for the former to precisely measure what is happening at the heart of the DLE subject to the study, and for the latter, to study its behavior in “real time”. But this purely descriptive approach can quickly be insufficient for those who want to understand the ins and outs of the phenomena observed. In fact, to address these issues in more detail, simultaneous quantitative and qualitative approaches, currently used in social sciences, become interesting insofar as they allow the researcher to be guided, offering him the ability to choose his own investigation targets (and not only those routinely provided by the platform). The tools at his disposal are traditional interviews, surveys and field observations. In this regard, we speculate that the model that represents the DLE at any given moment helps the modeler to identify and choose relevant investigation targets and that conversely, the data analysis results, which will be gathered and processed following this investigation, will be used to improve the DLE model in construction (iterative process).

Figure 5.30. IRIT simulation on the multi-agent Netlogo simulation platform. For a color version of this figure, see www.iste.co.uk/trestini/learning.zip

40 Presentation and platform download (in French) from the link: http://www.emse.fr/ ~picard/.

224

Modeling of Next Generation Digital Learning Environments

The analysis that traditionally follows these surveys, interviews and field observations is generally of a confirmatory character. They are based in large part on inferential statistics that aim to test, according to a hypothetico-deductive approach, formulated hypotheses with the help of tests such as the chi-squared, Fisher or student’s t tests, to name but a few. They use the theory of probabilities to restrict the number of individuals by allowing surveys on samples. Hypothesis tests then allow for decision-making in situations involving a degree of chance. They are particularly useful when it comes to crossing several variables between them and to determining the level of significance of their functional dependencies. These socalled multivariate analyses function well when there are not too many variables to deal with and when you can easily make assumptions about their behavior in the system. But as soon as one approaches the issue of highly complex systems, we are in the presence of so many variables and individuals (this not only in an MOOC) that these methods in turn show their limits. In addition, regarding phenomena that are often new and seldom documented (new generation DLE), hypotheses are difficult to formulate. Actually, without completely prohibiting a few variable crossovers when it seems useful and opportune, we need to consider other analysis methods to deal with complexity, because as John Wilder Tukey jokingly said at the beginning of 1970s, an approximate answer to the right problem is worth a good deal more than an exact answer to an approximate problem41. 5.5.3. Data analysis Once the data is collected, it is a matter of conducting the analysis itself. Facing complexity and interdependence, models are constructed based on the analysis we perform on data provided by the system, allowing the modeler to make assumptions on their overall behavior. This leads the modeler to experience “a model or design principles that can be used in the context of similar experiences” [HAR 09, p. 106]. We suggest using exploratory data analysis42 (EDA) since it relates to a complex system built up against descriptive statistics in order to deal with data retrieved from the different sources cited previously, namely: – dashboards, tracking tools; – simulation; 41 “It is better to be approximately right than exactly wrong” [LAD 97]. 42 The expression “data analysis” is to be taken here in a very general sense, and not in its smaller meaning referring to factor and classification analysis [LAD [97].

Modeling a DLE Perceived as a Complex System

225

– questionnaires and interviews. At the beginning of model construction, data analysis collected cannot only be exploratory (Tukey [TUK 77] quoted by Cabassut [CAB 12]), but it should also eventually make hypotheses which may be validated or invalidated by a confirmatory analysis. This uses inferential statistics to test hypotheses in a deductive approach. “[But] the practice rarely meets this scheme: you sometimes look to answer less specific questions, the data ‘pre-exists’ and has been collected according to complex surveys, there are absurd observations...” In these cases, their underlying hypotheses being violated, no longer have optimal classic methods. On the contrary, exploratory analysis starts from data and is based on observation logic. Thanks to a well-stocked toolbox, the explorer will look at his data from all sides, try to highlight structures, and, when appropriate, formulate plausible hypotheses” [LAD 97, p. 3]. But it would be absurd, as the author points out, to oppose classic statistics with exploratory statistics and equally absurd to want to confuse them. “They both occupy a place of choice in statistical analysis and can prove to be complementary” [LAD 97, p. 4]. Thus, for us, exploratory data analysis developed particularly by Tukey [TUK 77], backed by descriptive statistics, appeared to be a good choice for processing data analysis from a complex system. We have since adopted it for our research, especially since “it can help in the construction of a model” [LAD 97, p. 4]. Exploratory data analysis (EDA) uses an inductive approach to describe the population (here the different actors who have, one way or another, participated in the DLE) and move around the formulation of hypotheses at the end of research. Exploratory analysis first aims to familiarize us with new phenomena, little or non-documented, and thus helps us make hypotheses for future research. The hypothesis is then at the end of research, and not at the beginning, as it is obligatory in a confirmatory analysis. Furthermore, descriptive statistics that accompany and reinforce this exploratory analysis allows it to represent and synthesize numerous amounts of data provided by the system. “Descriptive statistics is the branch of statistics that brings together the

226

Modeling of Next Generation Digital Learning Environments

many techniques used to ‘describe’ a relatively important grouping of data”43. And as pointed out by Cabassut and Villette [CAB 12], “it has the advantage of avoiding the constraints of sample representativeness”. Taking the example of teachers participating in a course, they remind us that they are not necessarily a representative sample of the teacher population. “The institutional conditions vary greatly from one country to another; voluntary participant or participant designated by the school authority; training during working hours with a replacement professor or internship out of working time; an internship taken into account or not to advance career wise, etc” [CAB 12, p. 4]. For Joël Rosnay [DER 75, p. 122], continued by Tardieu et al. [TAR 86, p. 34], “system, modeling and simulation analysis constitute the three fundamental steps in the study of the dynamic behavior of complex systems” [TAR 86, p. 34]. During these past two years, we have engaged in a systemic modeling of the complexity approach in order to specifically study the behavior of several relatively popular MOOCs (project management, increasing MOOC from A to Z, EFAN and EFAN Math, etc.) seen as the latest generation DLEs. This work gave rise to four major publications: – a book on social MOOCs in France [TRE 16a], prefaced by Catherine Mongenet, a FUN project leader and resulting in the work of several researchers that we’ve directed within a seminar created for this purpose; – an article [TRE 15a] in the International Journal of Technologies in Higher Education, aiming to learn the perception of French actors in online teaching on this issue of MOOCs (which follows an announcement at the Montreal international conference (in Quebec); – an article [TRE 15b] published in The Online Journal of Distance Education and e-Learning, which presents our applied theoretical approach to “MOOCs in the paradigm of systemic modelling of complexity: some emerging properties”. – an article [TRE 17] presenting the modeling approach applied to help and support mechanisms offered in an MOOC. The publication is available on the journal website Distance et Médiatisation des saviors (DMS). This work has also opened the way for our future research program.

43 “Descriptive statistics”, Wikipedia (2018, updated May 13), available at: https:// en.wikipedia.org/wiki/Descriptive_statistics.

6 Modeling and Simulation of an MOOC: Practical Point

The modeling that we present below (which also partly serves to illustrate our approach) allows us to firstly provide a concrete illustration of what has been stated so far, and then to introduce our future research. In this chapter, we will therefore present this case before putting our mid-term research projects into perspective. Moreover, as this practical part serves as an example, it should be noted that we will follow the process of the systemic modeling of complexity developed in Chapter 5, by applying it to the specific case of a relatively well-known latest generation DLE: the Rémi Bachelet “Project management” MOOC (in its current version). Through the model which has been gradually constructed, our goal was to first obtain a visual representation of this instrumented activity in order to sufficiently understand the behavior. Another objective, common to all visual models, was accessing a communication aid between the different actors involved in this project, whether they are mere users or researchers. Thus, it has been possible to more easily share what we have learned concerning this device with the educational and scientific community. Among the different possible trajectories which leads us to systemic modeling (design, assessment, prospective analysis, etc.), we initially decided to limit ourselves to the analysis of system behavior in the hope of identifying some emerging properties (in a forward-looking approach). We have therefore supported, as recommended in section 6.5.2.4, a simulation program (in Python language) that we have written on the basis of this modeling1. Not having all the skills and energy necessary for the realization of this major project, we will only provide the founding 1 We will see how, technically, modeling helps the programmer to state the objects in its program.

Modeling of Next Generation Digital Learning Environments: Complex Systems Theory, First Edition. Marc Trestini. © ISTE Ltd 2018. Published by ISTE Ltd and John Wiley & Sons, Inc.

228

Modeling of Next Generation Digital Learning Environments

elements of this program, emphasizing the rigor that is required, and before considering a more important development. In doing so, we want to show that our recommendations are not only theoretical but they quickly find concrete applications. In the near future, our intention is obviously to continue the development of this project by using more significant and more expert human and material resources. 6.1. Modeling an MOOC Therefore, we wanted to model the “Project management” MOOC and attempt a simulation. Modeling an MOOC in its complexity involves first modeling a system of actions, in other words, identifying and formulating the functions performed by the system. We have done as described in section 6.1.1, bringing eight processors in a package diagram that we identified in section 6.1.4, corresponding to the eight constituent functions of any publicized educational environment. This diagram also represents the actors involved in each of the packages (Figure 6.1). We then created our use case diagrams which represent, let us recall, children processors and functionalities specifically necessary for users. In UML, it is possible to make a use case diagram for a MOOC diagram in its entirety or for each package. Here, for reasons of readability, we have decided to produce a use case for each identified processor: – awareness (Figure 6.2); – production (Figure 6.3); – meta-reflexive activities (Figure 6.4); – management and planning (Figure 6.5); – social interactions (Figure 6.6); – assessment (Figure 6.7); – information (Figure 6.8); – support (Figure 6.9).

Figure 6.1. Package diagram based on the eight constituent functions of any publicized educational environment

Modeling and Simulation of an MOOC: Practical Point 229

Modeling of Next Generation Digital Learning Environments

Figure 6.2. “Awareness” use case diagram

230

Modeling and Simulation of an MOOC: Practical Point

231

In addition, remember that these diagrams are unique because they represent an organization which itself is unique: that of a “Project management” MOOC. It certainly presents similarities with other MOOCs, but it also develops specific features that require a personalized study. It is in this sense that we wanted to put ourselves closer to the formulations chosen by the MOOC, rather than apply ours. So, we will call “students” the people we would have rather named “enrolled” in another context, etc. However, we had no trouble gathering our various action complexes in the eight constituent functions of any publicized educational environment, made evident by the following package and “use case” diagrams 6.1.1. Package and use case diagrams

Figure 6.3. “Production” use case diagram

232

Modeling of Next Generation Digital Learning Environments

Figure 6.4. Meta-reflexive activities use case

Figure 6.5. “Management and planning” use case

Modeling and Simulation of an MOOC: Practical Point

Figure 6.6. “Social interactions" use case diagram

233

Modeling of Next Generation Digital Learning Environments

Figure 6.7. “Assessment” use case

234

Figure 6.8. “Information” use case

Modeling and Simulation of an MOOC: Practical Point

235

236

Modeling of Next Generation Digital Learning Environments

Figure 6.9. “Support” use case

6.1.2. “Class” diagrams Once the system functions (or action complexes) are defined, we turned to the construction of class diagrams. Remember that this order is important in the systemic approach of complexity: it does not start by describing the system constituents, but its functions. Below, we give the project’s relatively representative class diagrams: – “Users” class (Figure 6.10); – “MOOC” class (Figure 6.11). Remember that the usefulness of class diagrams, from the perspective of a simulation, is that they allow the definition of different “objects” involved in the activity: what are they, what attributes do they have and what are the relationships and actions between these objects?

237

Figure 6.10. “Users” class diagram

Modeling and Simulation of an MOOC: Practical Point

Modeling of Next Generation Digital Learning Environments

Figure 6.11. MOOC class diagram

238

Modeling and Simulation of an MOOC: Practical Point

239

We have a concrete example which shows that object-oriented modeling (OOM) and programming of the same name are perfectly suited to the study of complex systems. To show this, we note that the “constituent” concept of a system which is linked to other constituents (or itself) by “dependence relationships” finds a concrete representation in UML in the “object” concept connected by “operations”. Moreover, to code this system in terms of “classes” makes the code independent of initial conditions and fluctuations from the environment (noise), which are transcribed in the script2 (main.py, Figure 6.15). This independence is assured by a clear separation between the object-oriented architecture that lists the parameters and defines the possible interactions, and that collects and initializes the parameters, calling upon operations, interactions or methods. That is in part what makes this type of programming flexible. Another flexibility factor lies in the fact that a class can be modified to affect all its instances. Finally, by comparing languages in systemic modeling (SM) and object-oriented modeling (OOM), it is interesting to note that: – “the system that ‘is’” in SM is “embodied” in OOM “attributes”; – “the system that ‘performs’” in SM is found in the OOM “methods”3 that act on other objects (and can make them evolve); – “the system that ‘becomes’” in SM is found in specific or external methods (feedback) in OOM which, respectively, act or react to the studied object. The object becomes the target of methods that, when applied to it, make it evolve (methods that belong to the object or not). OOM is very useful for modeling feedback. Calling a method in an object (initially, this is done in the script) generates a cascade of calls to various objects that can have a retroactive effect on the initial object for which the behavior will then be modified. 6.1.3. State transition diagrams We have also resorted to UML state transition diagrams that reflect, as their name suggests, the different states of processors: they characterize an evolving

2 The “script” is the file that contains executable statements that call upon “object methods” (“operations” in UML). 3 In object-oriented modeling (OOM) or in object-oriented programming (OOP), we’re talking about “methods”. These “methods” are called “operations” in UML.

240

Modeling of Next Generation Digital Learning Environments

system. State transition diagrams describe an object’s possible state changes in response to interactions with other objects or actors. As an example, below, we give the state transition diagram of a student enrolled in a “Project management” MOOC.

Figure 6.12. A student state transition diagram

6.1.4. Sequence diagrams Sequence diagrams allow for the representation of collaborations between objects from a temporal point of view. These diagrams do not describe an object’s state, they focus on the expression of interactions. They can also be used to illustrate a use case. The order to send a message is determined by its position on the vertical axis of the diagram: time flows “from top to bottom” on this axis. The layout of objects on the horizontal axis has no effect on the diagram’s semantics. State transition and sequence diagrams are the most important dynamic views in UML. As an example, we give two sequence diagrams: “general course” (Figure 6.13) and discussion thread (Figure 6.14).

241

Figure 6.13. “General course” sequence diagram

Modeling and Simulation of an MOOC: Practical Point

Modeling of Next Generation Digital Learning Environments

Figure 6.14. “Discussion thread” sequence diagram

242

Modeling and Simulation of an MOOC: Practical Point

243

6.2. MOOC simulation Simulation is fundamentally based on the previous systemic model. It concretely formulates in an object-oriented programming language: here in the Python language. Other languages of this kind could do the trick. The export of UML object attributes and operations to the Python program structure was performed by an export feature specific to the modeling software (Tools => Python => Generate code). Now, let us look at the program structure and the script that must be distinguished. 6.2.1. The program structure The program structure reflects the structure of the MOOC studied. Thus, the “Project management” MOOC is fundamentally composed of – a set of modules which themselves consist of a set of chapters; – a set of users “specializing”4 in students, teachers, tutors and referents. A student may specialize in “classic” or “advanced” and, if he has enough experience, in “student project by team”. A student is also defined by a number of attributes that will characterize his “learning” and “level of understanding” of the chapter, module or MOOC in progress. A student can “update” 5 (consisting of performing a series of operations including “read the course” and “answer the quiz”) and skip to the next chapter (with all modifications ensuing on his level of understanding). A chapter is composed of attributes (difficulty, associated teacher and associated quiz) and a method (view the course). If this method is called, it will also have a retroactive effect on the student (passed as a parameter) viewing the course and increasing his level of understanding of it, according to a law dependent on all parameters. A quiz is also an attribute (its difficulty) and a method (answering the quiz). This method also has an influence on the student (passed as a parameter): this shall be assigned a note that characterizes the student even further.

4 UML formulation, in the heritage sense in MOO. 5 “Updating” is an operation that involves a succession of operations.

244

Modeling of Next Generation Digital Learning Environments

Modeling and Simulation of an MOOC: Practical Point

245

Figure 6.15. Main.py script

6.2.2. The script The script takes place in three significant times as well as the temporal dimension of the simulation (Figure 6.15): – MOOC initialization (initialization du MOOC): learn about the “MOOC” object by assigning modules, chapters and a teacher; – registration phase (phase d’inscription): assign students to the MOOC (here, for example, 900 in a classic course and 100 in an advanced course); – learning phase (phase d’apprentissage): for every day of the course, for every student, “update” this (makes it read the course, answers the quiz, etc.). It is the “eleve.mettreAJour()” method that is responsible for this operation.

246

Modeling of Next Generation Digital Learning Environments

6.2.3. Related files Many other files are clearly associated with this script but are not immediately essential to the understanding of the program. We have decided to only provide a few as examples (see Appendices 3–5) to avoid unnecessarily overloading the book. 6.2.4. Parameters that come into play and initialization values In this example of simulation, the parameters that characterize the main objects are: – for the student: his commitment, competence, motivation, prerequisites, uniformly chosen at random between 0 and 1; – for the course: its difficulty, set at 0.5 by default; – for the teacher (there is only one): his pedagogy, his involvement, set to 1 by default. Note that if all stated parameters were at 1, the simulation predicts that the student would understand everything perfectly the first time. We have therefore assigned the student a random score following a normal distribution centered on his level of understanding with a standard deviation of 0.1. For example, if his level of understanding is 0.6, he will probably get 0.6 as a quiz score but will also be able to get a score of 0.5 or 0.7. 6.2.5. Progress in the chapters A student always tries the quiz twice to get the best possible score, unless he has obtained a score of 1 on the first attempt. If he has earned a score of 1 on the first quiz or has taken the quiz twice, he will proceed to the next chapter. Otherwise, he passes the quiz with a probability of 0.6. 6.3. Simulation result Would it be surprising to get the simulation results that correspond to nothing we expected to observe? Not really if one refers to the theories of SM developed thus far. According to the latter, we should rather expect to observe the unexpected phenomena and even counter-intuitive effects. And it is the latter that we have

Modeling and Simulation of an MOOC: Practical Point

247

witnessed. Indeed, wanting to study variations in the level of understanding of the course by students according to random initial conditions, we expected to obtain different levels of understanding. And against all odds, we have instead seen an amazing regularity in the system response (homeostasis). But let us make no mistake, and it is not so much this result that interests us in the first place, it is rather the implementation approach to achieve this result. That said, let us go over in detail regarding what an MOOC systemic modeling is and what its simulation allowed us to discover. One of the observable phenomena of the very simplified model presented here as an example is a representation of the students’ level of understanding. There are three levels: – the level of understanding of the chapter (niveauComprehensionChapitre), which corresponds to the understanding of the chapter of the course being studied; – the level of understanding of the module (niveauComprehensionChapitre), which corresponds to the understanding of the chapter of the course being studied; – the level of total understanding (niveauComprehensionTotale), which is the understanding of the entire MOOC. It should be noted that the level of understanding of the module is the average of the levels of understanding of all the chapters of the module and that the level of total understanding of the course is the average of all levels of understanding of all proposed chapters in the MOOC. In the sections that follow, we provide the average of all students of the three aforementioned variables based on time. 6.3.1. Level of total understanding The curve below is the average of all the MOOC’s levels of total understanding from all students according to the time (in days). We observe a growing understanding that slows toward the end of the MOOC. This slowdown is explained by the fact that the fastest students have already finished the MOOC and stop progressing.

248

Modeling of Next Generation Digital Learning Environments

6.3.2. Level of understanding of the module The curve shown in Figure 6.17 corresponds to the average of all the MOOC’s levels of understanding of the module from all students according to the time (in days).

Figure 6.16. The average of all levels of understanding of MOOC total of all students/day

Figure 6.17. The average of all levels of understanding of the module from all the students/day

Modeling and Simulation of an MOOC: Practical Point

249

Figure 6.18. The average of all levels of understanding of the module from all the students/day/random experience

The shape of this curve leads us to make four comments: – first observation: oscillations. This is explained by the fact that a student sees his level of understanding of the module evolving “unevenly”: he increases his understanding during the module, and when he moves to the next module, he begins with a void level of understanding. This is quite normal as he starts to appropriate new knowledge. But there is also an oscillation of the value of this quantity when considering the average of all students; – second observation: damped oscillations. The curve clearly shows damped oscillations (we can even venture to see an amplitude which decreases in exponential decreasing, etc.). This is explained by a progressive phase shift of students. At the beginning of the MOOC, all students start at the same time, which translates into a very fast growth, and then they are faced with a new module all at the same time, which translates into a very fast decline. But as students learn at different rates, some find themselves at the end of a module and therefore have a

250

Modeling of Next Generation Digital Learning Environments

high level of understanding, while others have already begun a new module and have a low level of understanding. In physics, this is called a dispersion phenomenon: the wave packet diminishes because the phase speeds of the different superimposed harmonic waves are different; – third observation: at the end of the MOOC, the level of understanding falls. We explain this phenomenon for technical reasons. When a student has finished the MOOC curriculum, he is no longer associated with a module, and therefore his level of understanding of the module remains void. This decline means that different students do not complete the MOOC simultaneously; ‒ fourth observation: we observe a certain regularity in the system response subject to variations in the initial conditions. In response to the completely random parameters supplied to the system, we can notice local fluctuations on the graph (especially in the damped part) when the initial conditions change (for different experiences), but the overall curve shape (damped oscillations) remains substantially the same. 6.3.3. Level of understanding of the chapter We observe a similar phenomenon to the level of understanding of the chapter.

Figure 6.19. The average of all levels of understanding of the chapter from all the students/day

Modeling and Simulation of an MOOC: Practical Point

251

Figure 6.20. The average of all levels of understanding of the chapter from all the students/day/random experience

At this point, it would be bold to immediately conclude the existence of any emergent phenomenon, revealing the stability of the system (its homeostatic nature) as long as it is subject to very different initial conditions. It was necessary for us to spend more time studying this regularity and try other tests, and this is not the desired goal here. Nevertheless, this result suggests the possibility of identifying, or even demonstrating, the recurrence of a behavior during different events (changes in the initial conditions) and allows for the revealing of the general causes and mechanisms to finally describe the different forms. 6.4. Putting the obtained results into perspective One might say that the study of a “comprehensionChapitre” variable has no great benefit because it does not reflect the actual level of a registered user. However, we can imagine that one registered user understands the course he is reading (therefore starting the course) very little and will tend to ask questions on a forum while a registered user understanding the course more fully (finishing the course) would be more independent. This could translate into forums that are oscillating in attendance

252

Modeling of Next Generation Digital Learning Environments

and, additionally, damped. It would have to be verified in the facts; but if that were proven, this result would be useful to possibly predict the sufficient number of tutors present on the forum to answer questions from registered users. Such a model, clearly expanded and adapted, can observe the unexpected phenomena, which can confirm or deny the field’s validity. But such models are superior to the simple observation, because they allow access to all the states of all participants at any moment, making the observation of any size possible. It is also important to remember that the result obtained was not very foreseeable, and we do not necessarily intuitively think that we should study the attendance of the forums over time... These are clearly just examples that may seem trivial compared to the modeling work that this requires. But nevertheless, they lead us to believe that this program (constructed according to the recommended modeling approach) was not used a lot (for lack of time and resources), while it already offers, such as it is, an interesting potential for the behavior analysis of Rémi Bachelet’s MOOC. It goes without saying that our goals are now to develop this model and its simulation program to produce further understanding about MOOCs and more generally the latest generation DLEs. Today, there are indeed a number of simulation techniques that allow for the observation of a complex system behavior to very varying degrees of abstraction and detail.

Conclusion

The purpose of this book is not to propose, once again, an improvement of the current DLE analysis models. It is rather to propose a new approach consisting of representing these environments, taking into account their complexity. We thus decided to abandon some past practices and use systems science and in particular the complex systems theory. As Bachelard said, access to science involves accepting the contradiction of the past. This epistemological rupture comes from our desire to break from the illusory belief that Cartesian and analytical modeling can reach the objectivity built in principle. Therefore, we decided to register the models on which we will reason in the paradigm of systemic modeling, one of the features of which is to consider models as living and evolutionary constructions rather than as preconceived data. Given the highly innovative nature of this change, our goal is not to just apply this systemic modeling to a particular environment (although practical applications are often very telling), but rather to cast the foundations of this new approach in order to initiate new research in this sense. Remember: this model has no vocation to serve to the design of an adaptive learning environment, which would claim to manage all the cases in order to allow personalized learning. So far, this goal seems impossible to achieve in view of the excessive amount of variables involved. The approach we propose instead seeks to model latest generation DLE using methods adapted to the infinite diversity of these variables, in the single hope of discovering some emerging properties or a few non-programmed adaptive behaviors. We will, however, always keep in mind that the complexity that characterizes these environments, and in particular their sensitivity to initial conditions, dismisses all hope to predict their behavior with certainty. However, is this degree of uncertainty a good reason to give up on this modeling project? Some Cartesian experimentalists will probably seek to move from this type of research to take interest in the study of objects that better lend themselves to the so-called “exact” sciences. The ensuing results will certainly be more rewarding. These experimentalists will have the satisfaction of producing knowledge, making it Modeling of Next Generation Digital Learning Environments: Complex Systems Theory, First Edition. Marc Trestini. © ISTE Ltd 2018. Published by ISTE Ltd and John Wiley & Sons, Inc.

254

Modeling of Next Generation Digital Learning Environments

possible to produce with certainty, at the same time ignoring other key issues which do not have this property. However, complex systems have become ubiquitous; the prediction of their behavior becomes almost obsessive to us humans beings. People want to know where they are going and what to expect. Predicting the weather in the long term, describing the trajectory of a leaf deposited on a stream, trying to regulate fishing worldwide, knowing the advantage of introducing digital tablets at a school or even predicting what MOOC will become in the future are just some examples of forward-looking or predictive research applicable to complex systems. And if the systemic modeling of complexity seems capable of producing interesting results, it is not unfortunately immune to errors or inaccuracies. Only the probability that an event may occur can sometimes be provided with certainty. In this regard, predictions remain uncertain. What approach should be taken? Initially, it is a matter of identifying the system boundaries to study. Thus, it is essential to consider a DLE (for example, a MOOC) as the union of the “digital environment” tool and all of its users. Indeed, assuming that a MOOC would only consist of a technical system, it would no longer concern a natural system of complexity, but simply a system of algorithmic complexity: this is the fundamental message of this book. We then break down the system (DLE) into subsystems: the “digital platform” system, its functional subsystems and “user” subsystems. It is essential to analyze each of these subsystems in terms of system modeling and avoid the pitfall of analytical modeling, at all costs. Rather than considering these subsystems as static organs of the system, whose behavior is independant from the other subsystems, we consider them instead as "black boxes" that accept inflow, emit outflow and change according to these flows and with time. To concretely model this in practice, it is advantageous to use the object concept in a computer sense. Thus, we create a “digital platform” class object and several objects all instantiated from a “user” class. Even before knowing the behaviors of these objects, the flow of each object through methods and attributes is defined. We therefore connect all these objects together, giving rise to the complex system skeleton being studied. Now, we come to the methodology that should be followed in order to analyze each of these subsystems. This varies greatly between the “digital platform” and “user” objects. The digital platform on which a MOOC operates is a system of algorithmic complexity. Its analysis is therefore tedious, but its behavior is entirely known. It is up to the modeler to define different objects that interact within this subsystem, that is, inflow and outflow and the behavior in response to this flow. This can be done with more or less detail according to the desired fineness1, but there is a maximum detail level, which exactly corresponds to the digital platform’s 1 For example, in an “object” forum, we can define the action, “Participate”, just as we can define more specific actions such as “Ask a question”, “Reply to a question”, “Mark a topic as resolved”, etc.

Conclusion

255

source code: this shows that it is an algorithmic complexity system. Thus, the modeler can make a user-robot interacting relationship (via the “user” object) with the original digital platform; but this approach, however, is not necessarily the most sensible because it would be necessary to code “user” objects that are complex enough to be able to interact with the digital platform, which is extremely difficult and quite possibly pointless. The user of a latest generation DLE, such as a MOOC, for example, is instead a system of natural complexity, and one can advance without too many risks that a human being is the most complex system that we know. Analytical methods must be different, because it is clearly impossible to consider the modeling of all actions of human activity. It is therefore necessary to restrict the diversity of human activity by only allowing the “user” object a limited number of stereotyped behaviors that have an important influence on the evolution of the general system and abandon the behaviors deemed as irrelevant in the context of the studied model. The modeler’s approach must follow a cycle of trial and error. First, they list all the parameters likely to influence the user’s behavior: these are the so-called active variables. They then look to obtain information concerning the relevance of these parameters, either by questioning users via a questionnaire, or by “spying” on their behavior in the DLE. Thanks to exploratory data analysis techniques (such as factorial correspondence analysis and, notably, the utilization of dendrograms), the modeler determines which parameters separate and which parameters do not. The so-called “dividing” parameters are those that allow us to create stereotypical user classes. The modeler can then determine which parameters to include in the modeling of the “user” objects and which settings to leave aside. At least two options are then offered: the modeller can either determine, through user classes, paragons on which they will base themself to code some “user” object types, either to give random values to separating settings of the “user” object according to a law of probability, reflecting the statistical data that has previously been identified. The second option was preferred in our study, as it allows for better observation of the sensitivity and resistivity of the system to changes in initial conditions. Once these subsystems have been coded, they must then be set in motion. A time dimension is added to the code by inserting a script2. We can observe the system’s behavior by observing the evolution of the variables that interest us. We can also repeat the modeling on several occasions to gauge the system’s sensitivity to random settings. The modeler can then compare the system’s different evolutions based on the values of the starting parameters. All this must be done first in order to check the consistency of the model: whether the behavior corresponds well enough to the observed reality of the DLE and whether the variables have been carefully chosen. 2 Part of the main code, this is the first one run by the computer, and it calls the different objects. In many object-oriented programming languages, the script is in the “hand” function.

256

Modeling of Next Generation Digital Learning Environments

Otherwise, they have either been poorly chosen or have not been modeled with sufficient accuracy in the DLE framework. The modeler then resumes their model in the light of all this data and enters a cycle of perpetual reassessment, until they obtain a model that reflects the observed reality. Once the model is considered coherent, they can finally use it for predictive purposes. Therefore, they can choose a particular population, initiate the simulation and observe the evolution of the system. They will be particularly attentive to the appearance in their model of unexpected phenomena. They can then alert decision-makers and question themself in the following way: “Given the population using the MOOC, is it possible that such a phenomena can happen?” But they will have to stay humble regarding the system complexity being studied and never say that a phenomenon, since it took place in their model, will necessarily take place in reality. In fact, the modeler must remain a scientist in all circumstances and never become a fortune-teller. The conclusions in this book may initially seem disappointing. We indeed renounce a fundamental objective in science: explaining a phenomenon completely and exhaustively. This is what we intended to do from CALL to ITS, with admittedly more or less success. But the more these technologies evolve, the more it seems irrefutable that these digital learning environments are ultimately complex systems. Yet, our mathematical models are unanimous: it is simply impossible to predict the evolution of such a system with certainty. It is, however, not necessary to stop at this defeatist statement, and this book tries to make use of it. Other scientific disappointments, like the fact that it is impossible to accurately predict the weather or even the development of a three-body system, have helped to break the deadlock of positivist determinism and to advance the sciences in other ways. Just as we are able to predict the weather for a few days, accepting that this forecast might be wrong, the transition from analytical modeling to systemic modeling will allow us to predict certain aspects of digital learning environment evolution by accepting the fact that what is presented is a possible evolution and not an absolute certainty. The price of this paradigm shift is that of scientific rigor: just as in they physical sciences, it will accurately define our uncertainties and determine the limits of our models in advance. In this way, and fortunately, our digital learning environment models will cease to be outdistanced by technopedagogical innovations; they can finally anticipate these innovations and predict some of their effects. We as researchers can thus guide our political decision-makers, support our educational engineers and confront the cognitive sciences to the reality of digital tools. For these reasons, we gladly accept to renounce analytical modeling for the benefit of system modeling; as Henry Ford said, “failure is simply an opportunity to begin again, this time more intelligently”3.

3 Source: Wikipedia, available at: https://en.wikiquote.org/wiki/Henry_Ford.

Appendices

Appendix 1 Functional model notation

Figure A1.1. A functional model notation of the OMT method described in [RUM 95]

Modeling of Next Generation Digital Learning Environments: Complex Systems Theory, First Edition. Marc Trestini. © ISTE Ltd 2018. Published by ISTE Ltd and John Wiley & Sons, Inc.

Appendix 2 Dynamic model notation

Figure A2.1. A dynamic model notation of the OMT method described in [RUM 95]

Modeling of Next Generation Digital Learning Environments: Complex Systems Theory, First Edition. Marc Trestini. © ISTE Ltd 2018. Published by ISTE Ltd and John Wiley & Sons, Inc.

Appendix 3 MOOC.py and Quiz.py

Figure A3.1. “MOOC-PY & Quiz” file – a screenshot from Pyzo software version 4.2.1

Modeling of Next Generation Digital Learning Environments: Complex Systems Theory, First Edition. Marc Trestini. © ISTE Ltd 2018. Published by ISTE Ltd and John Wiley & Sons, Inc.

Appendix 4 Étudiants.py

Figure A4.1(a). “Étudiants.py” file – a screenshot from Pyzo software version 4.2.1 Modeling of Next Generation Digital Learning Environments: Complex Systems Theory, First Edition. Marc Trestini. © ISTE Ltd 2018. Published by ISTE Ltd and John Wiley & Sons, Inc.

266

Modeling of Next Generation Digital Learning Environments

Figure A4.1(b). “Étudiants.py” – a screenshot from Pyzo software version 4.2.1 (continued)

Appendix 5 Chapitre.py

Figure A5.1. “Chapitre.py” file – a screenshot from Pyzo software version 4.2.1

Modeling of Next Generation Digital Learning Environments: Complex Systems Theory, First Edition. Marc Trestini. © ISTE Ltd 2018. Published by ISTE Ltd and John Wiley & Sons, Inc.

References

[ABR 94] ABRIC J.-C., “Les représentations sociales : aspects théoriques”, in ABRIC J.-C. (ed.), Pratiques sociales et représentations, PUF, Paris, 1994. [AMI 03] AMIGUES R., “Enseignement-apprentissage”, in AMIGUES R. (ed.), Petit vocabulaire raisonné à l’usage des professeurs débutants, available at: http://recherche.aix-mrs.iufm .fr/publ/voc/n1/amigues1/index.html (consulted on 23 October 2005), 2003. [ANC 92] ANCORI B., “Mémoire et apprentissage : de la neurobiologie à l’auto-organisation”, in ANCORI B. (ed.), Apprendre, se souvenir, décider. Une nouvelle rationalité de l’organisation, CNRS Editions, Paris, pp. 51–104, 1992. [ANC 05] ANCORI B., “Analogie, évolution scientifique et réseaux complexes”, Nouvelles perspectives en sciences sociales, no. 1, pp. 9–61, Prise de parole, Paris, available at: https://www.erudit.org/fr/revues/npss/2005-v1-n1-npss2908/602445ar.pdf, 2005. [ANI 06] ANIORTE P., CARIOU E., GOUARDERES E., “Modélisation de systèmes complexes distribués : l’ingénierie des modèles pour l’intégration des paradigmes “agent” et “composant””, Journée Multi-Agent et Composant (JMAC 2006), Nîmes, France, available at: http://csl.ensm-douai.fr/MAAC/uploads/anioriteJMAC2006.pdf, March 2006. [ART 90] ARTIGUE M., “Analyse de processus d’enseignement en environnement informatique”, Petit x, no. 26, pp. 5–27, available at: http://www-irem.ujf-grenoble.fr /revues/revue_x/fic/26/26x1.pdf, 1990. [AST 01] ASTOLFI J.-P., Piaget et Vygotsky : deux approches différentes du développement et de l’apprentissage, Éduquer et former, 2001. [ATL 79] ATLAN H., Entre le cristal et la fumée. Essai sur l’organisation du vivant, Le Seuil, Paris, 1979. [ATL 98] ATLAN H., “La fin du “tout génétique” ? Vers de nouveaux paradigmes en biologie”, Conférence-débat organisée par le groupe Sciences en questions, INRA, Paris, France, May 1998.

Modeling of Next Generation Digital Learning Environments: Complex Systems Theory, First Edition. Marc Trestini. © ISTE Ltd 2018. Published by ISTE Ltd and John Wiley & Sons, Inc.

270

Modeling of Next Generation Digital Learning Environments

[AUG 13] AUGUSTO V., Modélisation de systèmes complexes. Présentation de cours à l’ École nationale supérieure des mines de Saint-Etienne, available at: http://www.emse.fr/~augusto/enseignement/icm/gis1/UP2-1-Modelisation_Systemes_com plexes-handout.pdf, 2013. [AZU 97] AZUMA R., “A survey of augmented reality. Presence”, Teleoperators and Virtual Environments, vol. 6, no. 4, pp. 355–385, 1997. [BAC 00] BACKER M., “The roles of models in artificial intelligence and education research: a prospective view”, International Journal of Artificial Intelligence in Education, vol. 11, pp. 122–143, 2000. [BAK 96] BAK P., How Nature Works – The Science of Self-Organized Criticality, Springer Verlag, Berlin, 1996. [BAR 94] BARON M., “EIAO, quelques repères”, Terminal, technologie de l’information, culture et société, no. 65, available at: http://edutice.archives-ouvertes.fr/edutice00000297/fr/ (consulted on 17 May 2016), 1994. [BAR 04] BARBIER R., “Flash existentiel et reliance”, Le Journal des Chercheurs, available at: http://www.barbier-rd.nom.fr/journal/article.php3?id_article=148, 2004. [BAR 11] BARON G.-L., “Learning design”, Recherche et formation, no. 68, pp. 109–111, available at: http://rechercheformation.revues.org/1565 (consulted on 12 January 2015), 2011. [BAS 98] BASQUE J., DORÉ S., “Le concept d’environnement d’apprentissage informatisé”, Journal of Distance Education, Revue de l’éducation à distance, vol. 13, no. 1, pp. 40–56, available at: http://www.ijede.ca/index.php/jde/article/view/136/426 (consulted on 5 May 2015), 1998. [BAS 04] BASQUE J., “En quoi les TIC changent-elles les pratiques d’ingénierie pédagogique du professeur d’université ?”, Revue internationale des technologies en pédagogie universitaire, vol. 1, no. 3, pp. 7–13, available at: http://www.ritpu.org/IMG/pdf /basque.pdf (consulted on 25 March 2016), 2004. [BAT 77] BATESON G., Vers une écologie de l’esprit, Le Seuil, Paris, 1977. [BAT 00] BATAILLE M., “Représentation, implicitation, implication ; des représentations sociales aux représentations professionnelles”, in GARNIER C., ROUQUETTE M.L. (eds), Les représentations en éducation et formation, Éditions Nouvelles, Montreal, pp. 165–189, 2000. [BEA 10] BEAUNE A., “Théorie de l’Activité : applications au domaine des TICE”, Adjectif, available at: http://www.adjectif.net/spip/spip.php?article77, 10 October 2010. [BEG 01] BEGUIN P., RABARDEL P., Concevoir pour les activités instrumentées, available at: http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.472.2482&rep=rep1&type=pdf, 2001. [BEG 04] BEGUIN P., CERF M., “Formes et enjeux de l’analyse de l’activité pour la conception des systèmes de travail”, @ctivités, available at: http://activites.revues.org/1156 (consulted on 10 September 2016), April 2004.

References

271

[BEG 13] BEGUIN S., “Quelle différence entre complexe et compliqué ?”, Simplixi.fr, available at: http://www.simplixi.fr/difference-complexe-et-complique/, 10 January 2013. [BEH 09] BEHAZ A., DJOUDI M., “Approche de modélisation d’un apprenant à base d’ontologie pour un hypermédia adaptatif pédagogique”, in AMINE A., AIT MOHAMED O., ELBERICHI Z. (eds), Acte de la 2e Conférence Internationale sur l’informatique et ses Applications (CIIA’09), Saida, Algeria, available at: http://ceur-ws.org/Vol-547/130.pdf, May 2009. [BEJ 15] BEJAOUI R., PAQUETTE G., BASQUE J. et al., “Comment personnaliser l’apprentissage dans un cours en ligne ouvert et massif (CLOM) (MOOC) ?”, Actes du colloque Cirt@ 201, Université de Sherbrooke, available at: http://www.cirta.org/index.php/colloques/colloque-2015/44-actes-du-colloque-2015/186comment-personnaliser-l-apprentissage-dans-un-cours-en-ligne-ouvert-et-massif-clommooc-auteurs (consulted on 18 August 2015), 13 October 2015. [BEL 92] BELLEMAIN F., CAPPONI B., “Spécialité de l’organisation d’une séquence d’enseignement lors de l’utilisation de l’ordinateur”, Educational Studies in Mathematics, vol. 23, no. 1, pp. 59–97, 1992. [BEL 96] BELISLE C., LINARD M., “Quelles nouvelles compétences des acteurs de la formation dans le contexte des TIC ?”, Éducation Permanente, vol. 127, pp. 19–48, 1996. [BEN 06] BENAYED M., TRESTINI M., VERREMAN A., “Collaboration entre professeurs des écoles : intégration des TIC lors du stage en responsabilité”, in BARON G.-L., BRUILLARD E. (eds), Technologies de communication et formation d’enseignants : vers de nouvelles modalités de professionnalisation, INRP, Lyon, pp. 167–177, 2006. [BJØ 07] BJØRKLIA C.A., RØEDB B.K., BJELLANDB H.V. et al., “La théorie de l’activité comme aide à la conception d’équipement de navigation”, @ctivités, vol. 4, no. 1, pp. 170–178, available at: http://activites.revues.org/1471, 2007. [BLA 95] BLACKLER F., “Knowledge, knowledge work and organizations: an overview and interpretation”, Organization Studies, vol. 16, no. 6, pp. 1021–1046, 1995. [BLA 99] BLACKLER F., CRUMP N., MCDONALD S., “Managing experts and competing through innovation: an activity theoretical analysis”, Organization, vol. 6, no. 1, pp. 5–31, 1999. [BLA 13] BLANCO Á.F., GARCÍA-PEÑALVO F.J., SEIN-ECHALUCE M., “A methodology proposal for developing adaptive cMOOC”, Proceedings of the First International Conference on Technological Ecosystem for Enhancing Multiculturality ACM, pp. 553–558, 2013. [BLI 14] BLITMAN S., “Marcel Lebrun : “L’écart entre collaboration et aliénation numérique est étroit””, EducPros.fr, available at: http://www.letudiant.fr/educpros/entretiens/marcellebrun-l-ecart-entre-collaboration-et-alienation-numerique-est-etroit.html (consulted on 16 February 2016), 26 November 2014. [BOE 15] BOELAERT J., KHANEBOUBI M., “Une méthode de visualisation de traces d’activités sur la plateforme open edX : le cas du Mooc “Enseigner et former avec le numérique””, Adjectif, available at: http://www.adjectif.net/spip/spip.php?article350, 2015.

272

Modeling of Next Generation Digital Learning Environments

[BOG 81] BOGDANOV A., Essays in Tektology : The General Science of Organisation, translated by GORELIK G., Intersystem Publications, Seaside, CA, 1981. [BOU 00] BOURGUIN G., Un support informatique à l’activité coopérative fondé sur la théorie de l’activité : le sujet DARE, Doctoral thesis, University of Lille, available at: http://www-lisic.univ-littoral.fr/publis/1394700486.pdf, 2000. [BOU 04a] BOULDING K.E., “General systems theory, the skeleton of science”, E:CO Special Double Issue, vol. 6, nos 1–2, pp. 127–139, available at: http://pespmc1.vub.ac.be /books/Boulding.pdf, 2004. [BOU 04b] BOURGINE P., Vers une science des systèmes complexes, Communication presented at the Colloque de prospective, Paris, available at: http://cnsc.unistra.fr /uploads/media/CompteRenduCollque1.pdf, February 2004. [BRO 86] BROUSSEAU G., La théorisation des phénomènes mathématiques, Thesis, University of Bordeaux, 1986.

d’enseignement

des

[BRO 96] BRONCKART J.-P., Activité langagière, textes et discours: pour un interactionisme socio-discursif, Delachaux et Niestlé, Lausanne and Paris, 1996. [BRU 04] BRUILLARD E., “Apprentissage coopératif à distance : quelques repères sur les questions de recherché”, in SALEH I., BOUYAHI S. (eds), Enseignement à distance : épistémologie et usages, Hermès-Lavoisier Paris, pp. 115–135, available at: http://stef.ens-paris-saclay.fr/servlet/com.univ.collaboratif.utils.LectureFichiergw?CODE _FICHIER=1413450599933&ID_FICHE=34028, 2004. [BRU 06] BRUILLARD E., BARON G.-L., “Usages en milieu scolaire : caractérisation, observation et évaluation”, in GRANDBASTIEN M., LABAT J.-M. (eds), Environnements informatiques pour l’apprentissage humain, Hermès-Lavoisier, Paris, pp. 269–284, 2006. [BUR 05] BURGOS D., ARNAUD M., NEUHAUSER P. et al., “IMS Learning Design : la flexibilité pédagogique au service des besoins de la e-formation”, EpiNet – La revue électronique de l’EPI, available at: https://www.epi.asso.fr/revue/articles/a0512c.htm, 2005. [CAB 12] CABASSUT R., VILLETTE J.-P., “Un exemple d’analyse des croyances des enseignants envers l’enseignement de la modélisation”, in DORIER J.-L., COUTAT S. (eds), Enseignement des mathématiques et contrat social : enjeux et défis pour le 21ème siècle, Actes du colloque EMF, Geneva, pp. 668–677, 2012. [CAM 07] CAMBIEN A., Une introduction à l’approche systémique : Appréhender la complexité, Rapport du centre d’études sur les réseaux, les transports, l’urbanisme et les constructions publiques CERTU, Lyon, available at: http://lara.inist.fr bitstream /handle/2332/1431/CERTU-RE_08-09.pdf?sequence=1, 2007. [CHA 06] CHARLIER B. et al., “Apprendre en présence et à distance : une définition des dispositifs hybrides”, Distances et savoirs, vol. 4, no. 4, pp. 469–496, available at: http://archive-ouverte.unige.ch/unige:17649 (consulted on 26 December 2015), 2006. [CHE 10] CHERQUI-HOUOT I., TRESTINI M., SCHNEEWELE M., “Les usages d’un cahier de textes en ligne : Cas de l’ENT Lorrain, l’environnement PLACE”, Distances et Savoirs, vol. 8, no. 2, pp. 241–256, 2010.

References

273

[CHE 15a] CHEKOUR M., LAAFOU M., JANATI-IDRISSI R., “L’évolution des théories de l’apprentissage à l’ère du numérique”, EpiNet, no. 171, available at: http://www.epi .asso.fr/revue/articles/a1502b.htm (consulté le 26 December 2015), January 2015. [CHE 15b] CHEVRY E., ROSSINI I., TRESTINI M., “La notion d’aide en contexte numérique : le rôle des acteurs impliqués”, Actes de l’atelier de la conférence internationale EIAH 2015, Agadir, Morocco, available at: http://atelier-aide.sciencesconf.org/conference/atelier-aide /pages/actes_atelier_aide_EIAH2015.pdf, 2 June 2015. [CIS 12] CISEL M., BRUILLARD E, “Chronique des MOOC”, Rubrique de la Revue STICEF, no. 19, pp. 1–16, available at: http://sticef.univ-lemans.fr/num/vol2012/13r-cisel sticef _2012_cisel_13rp.pdf (consulted on 18 January 2016), 2012. [CIS 13] CISEL M., MOOC: comment tout a commencé. Dans La révolution MOOC, available at: http://blog.educpros.fr/matthieu-cisel/ (consulted on 15 April 2016), 3 December 2013. [CIS 16] CISEL M., “Les MOOC au sein de l’entreprise : avis aux DRH, La révolution MOOC”, available at: http://blog.educpros.fr/matthieu-cisel/ (consulted on 12 April 2016), 24 February 2016. [CLA 65] CLAUSIUS R., “Ueber verschidene für die Anwendung bequeme Formen der Hauptgleichungen der mechanischen warm theorie”, Annalen der Physic, vol.125, pp. 353-4100, 1865. [CLA 83] CLARK R.E., “Reconsidering research on learning from media”, Review Educational Research, vol. 53, no. 4, pp. 445–459, 1983. [CLA 91] CLARK R.E., CRAIG T.G., “Research and theory on multi-media learning effects”, in GIARDINA M. (ed.), Interactive Multimedia Learning Environments, Springer-Verlag, Berlin, pp. 19–30, 1991. [CLA 94] CLARK R.E., “Media will never influence learning”, Educational Technology Research and Development, vol. 42, no. 2, pp. 21–29, 1994. [CLA 09] CLARK R.E., “À la recherche des ingrédients actifs de l’apprentissage. Les effets des dispositifs d’enseignement à distance”, Distances et Savoirs, vol. 7, no. 1, p. 114, 2009. [CON 03] CONTAMINES J., GEORGE S., HOTTE R., “Approche instrumentale des banques de ressources éducatives”, Sciences et Techniques Educatives, vol. 10, Special issue, pp. 157–178, available at: https://halshs.archives-ouvertes.fr/hal-00298189/document, 2003. [CRA 01] CRAHAY M., “Psychologie de l’éducation : les fondements de l’apprentissage”, Éduquer et Former, Sciences Humaines, Auxerre, 2001. D., “Le connectivisme : une théorie socio-informatique [CRI 12] CRISTOL de l’apprentissage”, available at: http://4cristol.over-blog.com-article-le-connectivismeune-theorie-socio-informatique-de-l-apprentissage-113809666.html (consulted on 12 January 2016), 6 December 2012.

274

Modeling of Next Generation Digital Learning Environments

[DAM 08] DAMBREVILLE S.C., “Évaluer un dispositif de formation à distance : principes et retour d’expérience”, Revue des interactions humaines médiatisées, vol. 9, no. 2, pp. 25–52, 2008. [DEL 99] DE LANNOY J.D., FEYEREISEN P., Qu’est-ce-donc qu’apprendre ?, Delachaux et Niestlé, Paris, 1999. [DEL 00] DELANNOY C., “Un cercle vertueux “affectif – cognitive” ?”, Les cahiers pédagogiques, no. 381, February 2000. [DEM 08] DEMAIZIÈRE F., “Le dispositif, un incontournable du moment”, Alsi, vol. 11, no. 2, pp. 157–161, available at: http://alsic.revues.org/384, 2008. [DEN 13] DENNERY M., “Macro, méso et micro pédagogie: clarifions les definitions”, available at: http://www.blog-formation-entreprise.fr/?p=2643 (consulted on 30 June 2016), 30 July 2013. [DEP 01] DEPOVER C., “Réinventer la pédagogie, les nouveaux paradigmes éducatifs”, Actes du colloque Initiatives 2001, Éthique et nouvelles technologies, L’appropriation des savoirs en question, Beyrouth, Lebanon, available at: http://www.initiatives.refer.org /Initiatives-2001/_notes/sess501.htm, September 2001. [DEP 02] DEPOVER C., MARCHAND L., E-Learning et formation des adultes en contexte professionnel, De Boeck & Larcier, Brussels, 2002. [DEP 09] DEPOVER C., KARSENTI T., KOMIS V., Enseigner avec les technologies : favoriser les apprentissages, développer les compétences, Presses de l’Université du Québec, Quebec, 2009. [DER 75] DE ROSNAY J., Le macroscope : vers une vision globale, Le Seuil, Paris, 1975. [DES 99] DESSUS P., LEMAIRE B., “La technologie des médias joue-t-elle un rôle dans la transmission des connaissances ?”, in AGOSTINELLI S. (ed.), Comment penser la communication des connaissances ? Du CD-ROM à l’Internet, pp. 253–264, L’Harmattan, Paris, available at: http://webcom.upmf-grenoble.fr/sciedu/pdessus/media 99.pdf, 1999. [DES 03] DESMOULINS C., MARQUET P., BOUHINEAU D., “Introduction”, Actes de la conférence EIAH 2003, Strasbourg, France, 15–17 April 2003. [DES 06a] DESMOULINS C., GRANDBASTIEN M., “Une ingénierie des EIAH fondée sur des ontologies”, in GRANBASTIEN M., LABAT J.M. (eds), Environnements informatiques pour l’apprentissage humain, Hermès-Lavoisier, Paris, pp. 160–180, 2006. [DES 06b] DESSUS P., “Quelles idées sur l’enseignement nous révèlent les modèles d’Instructional Design ?”, Revue suisse des sciences de l’éducation, vol. 28, no. 1, pp. 137–157, available at: http://tecfaetu.unige.ch/etu-maltt/volt/venni6/bases/carte1 /SZBW_6.1_Dessus.pdf, 2006. [DES 09] DESSUS P., MARQUET P., “Éditorial”, Distances et savoirs, vol. 7, no. 10, 2009.

References

275

[DES 13] DESROSIERS C., Analyse de pratiques pédagogiques d’enseignantes et d’enseignants du réseau collégial recourant à des environnements numériques d’apprentissage en enseignement hybride et propositions de stratégies optimales d’utilisation, Thesis, Université de Sherbrooke, Canada, available at: https://savoirs.usherbrooke.ca /handle/11143/9670, 2013. [DEV 10] DEVELOTTE C., MANGENOT F., “Feedback correctifs dans des formations de tuteurs de langue en ligne (en synchrone et en asynchrone)”, Actes du congrès de l’Actualité de la recherche en éducation et en formation (AREF), University of Geneva, available at: https://plone.unige.ch/aref2010/symposiums-courts/coordinateurs-en-e/professionnalite-d es-enseignants-se-former-dans-les-environnements-numeriques-2019apprentissage/Feedback%20correctifs.pdf (consulted on 18 August 2015), September 2010. [DIR 16] DIR M., SIMONIAN S., “Analyse de l’activité tutorale dans un organisme privé de formation à distance selon le modèle d’Engeström : tensions et écarts au prescrit”, Distances et mediations des saviors, no. 11, available at: http://dms.revues.org/1109, 2016. [DOI 81] DOISE W., MUGNY G., Le développement social de l’intelligence, InterÉditions, Paris, 1981. [DOI 86] DOISE W., “Les représentations sociales : définition d’un concept”, in DOISE W., PALMONARI A. (eds), L’étude des représentations sociales, Delachaux et Niestlé, Lausanne, Paris, pp. 81–94, 1986. [DOI 97] DOISE W., MUGNY G., Psychologie sociale et développement cognitive, Armand Colin, Paris, 1997. [DOM 10] DOMINGUES C., Interaction 3D Collaborative en Réalité Virtuelle, PhD Thesis, University of Evry, Val d’Essonne, available at: https://tel.archives-ouvertes.fr/tel00542170/document, 2010. [DON 02] DONNADIEU G., KARSKY M., La systémique, penser et agir dans la complexité, Éditions de liaisons, Paris, 2002. [DON 03] DONNADIEU G., DURAND D., NEEL D. et al., L’Approche systémique : de quoi s’agit-il ?, Synthèse des travaux du Groupe AFSCET “Diffusion de la pensée systémique”, available at: http://www.afscet.asso.fr/SystemicApproach.pdf, 2003. [DUC 05] DUCROCQ-HENRY S., “Le “Lan Party” : Aspects de socialisation, d’identité et d’apprentissage reliés au jeu vidéo en réseau”, Conférence au département d’Éducation de l’UQAT, available at: http://www.ludisme.com/media/Conferen-ce_SDH_LAN.pdf, 16 June 2005. [DUF 14] DUFOUR H., “La classe inversée”, Technologie, no. 193, pp. 44–47, available at: http://eduscol.education.fr/sti/sites/eduscol.education.fr.sti/files/ressources/techniques/650 8/6508-193-p44.pdf (consulted on 12 March 2017), September 2014. [END 12] ENDRIZZI L., “Les technologies numériques dans l’enseignement supérieur, entre défis et opportunités”, Veille et Analyses, no. 78, available at: http://ife.enslyon.fr/vst/DA/detailsDossier.php?parent=accueil&dossier=78&lang=fr (consulted on 16 February 2004), October 2012.

276

Modeling of Next Generation Digital Learning Environments

[ENG 87] ENGESTRÖM Y., Learning by Expanding: An Activity-Theoretical Approach to Developmental Research, Orienta-Konsultit Oy, Helsinki, 1987. [ENG 00] ENGESTRÖM Y., “Activity theory as a framework for analysing and redesigning work”, Ergonomics, vol. 43, no. 7, pp. 960–974, 2000. [ENG 05] ENGESTRÖM Y., Developmental Work Research: Expanding Activity Theory into Practice, Behmanns Media, Berlin, 2005. [FAB 13] FABRE I., “L’espace documentaire comme lieu de médiations”, Archives ouvertes HAL, available at: https://hal.archives-ouvertes.fr/hal-00837950/document, June 2013. [FER 05] FERRARIS C., LEJEUNE A., VIGNOLLET L. et al., “Modélisation de scénarios pédagogiques collaboratifs”, Archives ouvertes HAL, available at: https://hal.archives-ouv ertes.fr/hal-00005687/document, June 2005. [FOR 05] FORTIN R., Comprendre la complexité. Introduction à la Méthode d’Edgar Morin, Presses de l’université Laval, Quebec, 2005. [GAG 68] GAGNE R., “Learning categories and instructional strategies”, La recherche en enseignement programmé : tendances actuelles, Actes d’un colloque OTAN, Dunod, Paris, pp. 155–170, 1968. [GIL 95] GILLY M., “Approches socio-constructives du développement cognitif”, in GAONACH G., GOLDER C. (eds), Manuel de psychologie pour l’enseignement, Hachette, Paris, pp. 130–167, 1995. [GIO 01] GIORDAN A., “(Re)construire les connaissances”, in RUANO-BORBALAN J.C. (ed.), Eduquer et former : Les connaissances et débats en éducation et en formation, Éditions Sciences Humaines, Paris, pp. 95–101, 2001. [GLE 87] GLEICK J., Chaos: The Making of a New Science, Penguin Books, New York, 1987. [GOG 16] GOGUEL J., “Modèle. Les modèles dans la science de la Terre”, in BOUDON R., DAMISCH H., GOGUEL J. et al. (eds), Les grands articles, Encyclopaedia Universalis, Paris, available at: http://www.universalis.fr/encyclopedie/modele/, 2016. [GON 04] GONZALEZ C., Le rôle de l’apprentissage mixte dans le monde de la technologie, Benchmarks Online, University Information Technology, available at: http://www.unt.edu/benchmarks/archives/2004/september04/eis.htm (consulted on 12 January 2016), 2004. [GRA 03] GRANDBASTIEN M., OUBAHSSI L., CLAËS G., “A process oriented approach for modelling on line Learning Environments”, Intelligent Management Systems, AIED 2003 Supplemental Proceedings, no. 4, pp. 140–152, University of Sydney, 2003. [GUI 04] GUITÉ F., Le connectivisme (néo socioconstructivisme.), available http://www.francoisguite.com/2004/12/le-connectivisme-neo-socioconstructivisme/ (consulted on 17 January 2016), December 2004.

at:

[HAR 01] HARVEY P.L., LEMIRE G., La nouvelle éducation : NTIC, Transdisciplinarité et Communautique, Presses de l’université Laval, Quebec, 2001.

References

277

[HAR 09] HARVEY S., LOISELLE J., “Proposition d’un modèle de recherche développement”, Recherches qualitatives, vol. 28, no. 2, pp. 95–117, available at: http://www.recherchequalitative.qc.ca/documents/files/revue/edition_reguliere/numero28 (2)/harvey(28)2.pdf, 2009. [HEN 01] HENRI F., LUNDGREN-CAYROL K., Apprentissage collaboratif à distance : Pour comprendre et concevoir les environnements d’apprentissage virtuels, Presses de l’université du Québec, Sainte-Foy, 2001. [HEN 07] HENRI F., COMPTE C., CHARLIER B., “La scénarisation pédagogique dans tous ses débats…”, Revue internationale des technologies en pédagogie universitaire, vol. 4, no. 2, pp. 14–24, available at: http://www.ritpu.org/IMG/pdf/ritpu0402_henri. pdf, 2007. [HEN 14] HENNING P.A., HEBERLE F., STREICHER A. et al., “Personalized web learning: merging open educational resources into adaptive courses for higher education”, Personalization Approaches in Learning Environments, pp. 55–62, 2014. [HOU 88] HOUSSAYE J., Le triangle pédagogique, Peter Lang, Berne, 1988. [JAC 93] JACQUINOT G., “Apprivoiser la distance et supprimer l’absence ? ou les défis de la formation à distance”, Revue Française de Pédagogie, no. 102, pp. 55–67, January–February–March 1993. [JAR 03] JARZABKOWSKI P., “Strategic practices: an activity theory perspective on continuity and change”, Journal of Management Studies, vol. 40, no. 1, p. 22, 2003. [JEN 09] JENNI J.C., Objets d’apprentissage, ressources pédagogiques : le point de vue d’enseignants universitaires sur les pratiques de mutualisation et d’échanges : une analyse exploratoire des conceptions d’enseignants du supérieur, Thesis, University of Geneva (TECFA), 2009. [JER 96] JERMANN P., Conception et analyse d’une interface semi-structurée dédiée à la co-résolution de problème, Thesis, University of Geneva, available at: http://vcampus.uom.ac.mu/orizons/html/res270704/CSCL-CSCW/CoRes.unige/~jermann /staf/colin-1.html (consulted on 11 September 2016), 1996. [JON 00] JONASSEN D.H., “Revisiting activity theory as a framework for designing student-centered learning environments”, in JONASSEN D.H., LAND S.M. (eds), Theoretical Foundations of Learning Environments, Lawrence Erlbaum, Mahwah, pp. 81–121, 2000. [JON 02] JONASSEN D.H., “Technology as cognitive tools: learners as designers”, ITForum, Pennsylvania State University, PA, available at: http://citeseerx.ist.psu.edu/viewdoc /download?doi=10.1.1.467.4199&rep=rep1&type=pdf, 2002. [JON 09] JONASSEN D.H., Handbook of Research for Educational Communications and the Technology, Simon & Schuster Macmillan, New York, 2009. [KAD 16] KADIO J., PREVOST L., REMMER M., Jouons à apprendre à connaître les solides à l’aide de la Réalité Augmentée, Projet d’apprentissage de notions géométriques ponctué d’activités relevant de la réalité augmentée, Travail de recherche produit dans le cadre de l’UE 34 “Maîtrise et usages du e-Leraning pour la formation” du Master CFT, University of Strasbourg, 2016.

278

Modeling of Next Generation Digital Learning Environments

[KAR 01] KARSENTI T., LAROSE F., Les TIC… au cœur des pédagogies universitaires, Presses de l’université du Québec, Quebec, 2001. [KOP 00] KOPER R., From Change to Renewal Educational Technology Foundations of Electronic Learning Environments, Open University of the Netherlands, Heerlen, available at: http://dspace.ou.nl/bitstream/1820/38/2/koper-inaugural-address-eng.pdf (consulted on 10 July 2016), 2000. [KOP 01] KOPER R., Modeling Units of Study from a Pedagogical Perspective – The Pedagogical Metamodel Behind EML, Open University of the Netherlands, Heerlen, available at: http://dspace.ou.nlbitstream/1820/36/1/Pedagogical%20metamodel%20 behind%20EMLv2.pdf (consulted on 2 July 2016), 2001. [KOP 03] KOPER R., “Combining re-usable learning, resources and services to pedagogical purposeful units of learning”, in LITTLEJOHN A. (ed.), Reusing Online Resources: A Sustainable Approach to eLearning, Kogan Page, London, pp. 46–59, 2003. [KOZ 91] KOZMA R.B., “Learning with media”, Review Educational Research, vol. 6, no. 2, pp. 179–211, 1991. [KOZ 94] KOZMA R.B., “Will media influence learning? Reframing the debate”, Educational Technology Research and Development, vol. 42, no. 2, pp. 7–19, 1994. [KOZ 05] KOZANITIS A., Les principaux courants théoriques de l’enseignement et de l’apprentissage : un point de vue historique, Polytechnic School of Montreal, available at: http://www.polymtl.ca/bap/docs/documents/historique_approche_en-seignement.pdf (consulted on 27 December 2006), September 2005. [KUU 96] KUUTTI K., “Activity theory as a potential framework for human–computer interaction research”, in NARDI B. (ed.), Context and Consciousness: Activity Theory and Human Computer Interaction, MIT Press, Cambridge, pp. 17–44, 1996. [LAD 97] LADIRAY D., “L’AED, analyse exploratoire des données”, Courrier des statistiques, pp. 83–84, available at: http://www.insee.fr/fr/ffc/docs_ffc/cs 90a.pdf (consulted on 12 April 2016), 1997. [LAF 05a] LAFORCADE P., “Approche par transformation de modèles pour la conception d’EIAH : Illustration entre les langages CPM et IMS-LD”, in TCHOUNIKINE P., JOAB M., TROUCHE L. (eds), Actes du colloque Environnements Informatiques pour l’Apprentissage Humain (EIAH), Montpellier, pp. 213–224, available at: https://halshs.archives-ouvertes .fr/hal-00005664/document, 2005. [LAF 05b] LAFORCADE P., NODENOT T., SALLABERRY C., “Un langage de modélisation pédagogique basé sur UML”, Revue STICEF, no. 12, available at: http://sticef.univlemans.fr/num/vol2005/laforcade-05/sticef_2005_laforcade_05p.pdf, 2005. [LAG 10] LAGUECIR A., COLAS H., MASSUE M.-L., “Des modes d’utilisation du plan d’affaires : apports de la théorie de l’activité”, Revue de l’Entrepreneuriat, vol. 1, no. 9, pp. 50–81, available at: https://www.cairn.info/load_pdf.php?ID_ARTIC LE=ENT RE_091_0004, 2010.

References

279

[LEB 02] LEBRUN M., Théories et méthodes pour enseigner et apprendre : quelle place pour les TIC dans l’éducation?, De Boeck, Brussels, 2002. [LEC 06] LECAS G., “La modélisation systémique : outils méthodologiques pour économistes”, Innovations, vol. 2, no. 24, pp. 199–230, available at: https://www.cairn .info/revue-innovations-2006-2-page-199.htm, 2006. [LEF 12] LEFEVRE M., BROISIN J., BUTOIANU V. et al., “Personnalisation de l’apprentissage : comparaison des besoins et approches à travers l’étude de quelques dispositifs”, STICEF, no. 19, pp. 353–387, available at: http://sticef.univ-lemans.fr/num/vol 2012/06-lefev reindividualisation/sticef_2012_NS_lefevre_06.htm, 2012. [LEG 93] LEGENDRE R., Dictionnaire actuel de l’éducation, Guérin, Montreal, 1993. [LEG 04] LEGENDRE M.-F., “Approches constructivistes et nouvelles orientations curriculaires : d’un curriculum fondé sur l’approche par objectifs à un curriculum axé sur le développement de compétences”, in JONNAERT P., MASCIOTRA D. (eds), Constructivisme, choix contemporains : hommage à Ernst Von Glasersfeld, Presses de l’université du Québec, Sainte-Foy, Quebec, pp. 51–87, 2004. [LEI 70] LEIF J., RUSTIN G., Philosophie de l’éducation : Pédagogie générale, vol. 1, Delagrave, Paris, 1970. [LEM 77] LE MOIGNE J.-L., La théorie du système général : Théorie de la modélisation, PUF, Paris, available at: http://www.mcxapc.org/inserts/ouvrages/0609tsgtm.pdf, 1977. [LEM 94] LE MOIGNE J.-L., La théorie du système général : Théorie de la modélisation, PUF, Paris, available at: http://www.mcxapc.org/inserts/ouvrages/0609tsgtm.pdf, 1994. [LEM 99] LE MOIGNE J.-L., La modélisation des systèmes complexes, Dunod, Paris, 1999. [LEM 08] LEMIRE G., Modélisation et construction des mondes de connaissances : Aspects constructiviste, socioconstructiviste, cognitiviste et systémique, Presses de l’université Laval, Québec, 2008. [LÉO 76] LÉONTIEV A.N., Le développement du psychisme, Éditions Sociales, Paris, 1976. [LÉO 78] LÉONTIEV A.N., Activity, Englewood Cliffs, 1978.

Consciousness,

and

Personality,

Prentice-Hall,

[LÉO 81] LÉONTIEV A.N., “The problem of activity in psychology”, in WERTSCH J.V. (ed.), The Concept of Activity in Soviet Psychology, M.E. Sharpe Inc., Armonk, 1981. [LEP 15] LEPAGE D., Moodle Ajout d’un Scorm en mode “Manifeste SCORM externe”, available at: http://docplayer.fr/4480373-Moodle-2-8-3-ajout-d-un-scorm-en-modemanifeste-scorm-externe.html, 14 February 2015. [LEW 98] LEWIS R., “Learning together: a rationale, some experiences and a framework”, in ROUET J.-F., DE LA PASSARDIERE B. (eds), Actes du 4e colloque Hypermédias et Apprentissages, Poitiers, France, 15–17 October 1998.

280

Modeling of Next Generation Digital Learning Environments

[LHE 10] LHEUREUX F., “Représentations professionnelles, satisfaction au travail et choix de carrière des personnels infirmiers : le rôle des valeurs d’autonomie”, Psychologie du travail et des organisations (PTO), vol. 16, no. 4, pp. 312–324, available at: http://revuepto.com/articles%20pdf/Decembre%202010/Vol%2016-4-2.pdf, 2010. [LIN 02] LINARD M., “Conception de dispositifs et changement de paradigme en formation”, Les TIC au service des nouveaux dispositifs de formation, Education permanente, no. 152, 2002. [MAM 11] MAMMERI Z., Modélisation du comportement du système en UML, Cours de M2P GLRE (Génie Logiciel, logiciels Répartis et Embarqués), Paul Sabatier University, Toulouse III, available at: http://www.irit.fr/~Zoubir.Mammeri/Cours/M2PConcSTR /Chap4UML.pdf (consulted on 10 October 2015), 2011. [MAN 09] MANDELBROT B., Fractales, hasard et finance, Flammarion, Paris, 2009. [MAR 04] MARTEL C., FERRARIS C., CARON B. et al., “A model for CSCL allowing tailorability: implementation in the electronic schoolbag groupware”, Proceedings of the 10th CRIWG Conference, LNCS 3198, San Carlos, Costa Rica, 2004. [MAR 05] MARQUET P., “Intérêt du concept de conflit instrumental pour la compréhension des usages des EIAH”, in TCHOUNIKINE P., JOAB M., TROUCHE L. (eds), Actes de la conférence EIAH 2005, INRP, Institut Montpellier II, 2005. [MAR 11] MARQUET M., “eLearning et conflit instrumental : entre didactique, pédagogie et technique”, Recherche et Formation, no. 68, pp. 31–46, available at: http://recherche formation.revues.org/1499 (consulted on 6 December 2015), 2011. [MAR 12] MARTIN P., CLOUVEL P., REITZ P. et al., Modélisation d’un système complexe : une méthode declarative, Report, 13005, LIRMM, available at: https://hal.archivesouvertes.fr/hal-00706573/document, 2012. [MAR 13] MARTINEAU P., BARMA S., Comprendre l’appropriation du iPad comme outil d’enseignement et d’apprentissage dans une école secondaire en milieu défavorisé dans le cadre du projet FAST, Communication presented at ISCAR C-US, Interactive Symposium, available at: http://crires.ulaval.ca/sites/crires/files/2013/2-1335martineau.pdf, July 2013. [MAR 15] MARCEL F., SPRIET T., “Un tableau de bord pédagogique pour MOODLE”, available at: http://moodlemoot2014.univ-paris3.fr/course/view.php?id=211 (consulted on 17 November 2017), June 2015. [MAZ 97] MAZUR E., Peer Instruction: A User’s Manual, Prentice Hall, 1997. [MEI 94] MEIRIEU P., Apprendre… oui, mais comment ?, Éditions ESF, Paris, 1994. [MEL 10] MELHUISH K., FALLOON G., “Looking to the future: M-learning with the iPad. Computers”, New Zealand Schools: Learning, Leading, Technology, vol. 22, no. 3, available at: http://www.otago.ac.nz/cdelt/otago064509.pdf, 2010. [MEU 04] MEUNIER J.P., PERAYA D., Introduction aux théories de la communication. Analyse sémio-pragmatique de la communication médiatique, De Boeck, Brussels, 2004.

References

281

[MIT 09] MITCHELL M., Complexity, Guided Tour (349), Oxford University Press, Oxford, 2009. [MŒG 98] MŒGLIN P., L’industrialisation de la formation. État de la question, NDP, Paris, 1998. [MOR 73] MORIN E., Le paradigme perdu : la nature humaine, Le Seuil, Paris, 1973. [MOR 90] MORIN E., Science avec Conscience, Fayard, Paris, 1990. [MOR 08] MORIN E., La Méthode, Le Seuil, Paris, 2008. [MOU 15] MOUDNI S., Impact du tutorat sur le taux de réussite des MOOC, Thesis, University of Strasbourg, 2015. [MUS 07] MUSSO P., PONTHOU L., SEULLIET É., Fabriquer le futur 2. L’imaginaire au service de l’innovation, Village Mondial, Paris, 2007. [NIO 77] NIOCHE J.-P., “Pour un bon usage de l’expérimentalisme dans l’évaluation des méthodes pédagogiques”, Enseignement et gestion, no. 2, pp. 69–86, 1977. [NIO 82] NIOCHE J.-P., “De l’évaluation à l’analyse des politiques publiques”, Revue française de science politique, vol. 32, no. 1, pp. 32–61, available at: http://www.persee.fr /doc/rfsp_0035-2950_1982_num_32_1_393996, 1982. [NOD 07a] NODENOT T., LAFORCADE P., LE PALLEC X., “Visual design of coherent technology-enhanced learning systems: a few lessons learnt from CPM language”, in BOTTURI L., STUBBS S.T. (eds), Handbook of Visual Languages in Instructional Design: Theories and Practices, IDEA Group, Hershey, pp. 254–280, 2007. [NOD 07b] NODENOT T., “Scénarisation pédagogique et modèles conceptuels d’un EIAH : que peuvent apporter les langages visuels ?”, Revue internationale des technologies en pédagogie universitaire, vol. 4, no. 2, pp. 85–102, 2007. [NON 00] NONGA HONLA J., LE MOIGNE J.-L., “La Modélisation des systèmes complexes”, Les fiches de lecture de la Chaire D.S.O., available at: http://mip-ms.cnam.fr/servlet /com.univ.collaboratif.utils.LectureFichiergw?ID_FICHIER=1295877018078, 2000. [OUB 05] OUBAHSSI L., GRANDBASTIEN M., NGOMO M. et al., “The activity at the center of the global open and distance learning process”, The 12th International Conference on Artificial Intelligence in Education (AIED 2005), Amsterdam, pp. 386–394, 2005. [OUB 07] OUBAHSSI L., GRANDBASTIEN M., “Une généralisation du modèle d’activité d’IMS LD pour les systèmes de e-formation”, E-TI : la revue électronique des technologies de l’information, no. 4, available at: http://5.135.72.12/document.php?id= 1476 (consulted on 30 September 2017), 2007. [OUZ 11] OUZILOU O., Rationalité épistémique et sciences sociales, Aix-en-Provence, available at: https://halshs.archives-ouvertes.fr/hal-00664684/, 2011. [PAP 81] PAPERT S., Jaillisement de l’esprit : ordinateurs et apprentissage, Flammarion, Paris, 1981.

282

Modeling of Next Generation Digital Learning Environments

[PAQ 02a] PAQUETTE G., Modélisation des connaissances et des compétences, Presses de l’université du Québec, Quebec, 2002. [PAQ 02b] PAQUETTE G., L’Ingénierie pédagogique : pour construire l’apprentissage en réseau, Presses de l’université du Québec, Quebec, 2002. [PER 83] PERRIAULT J., “Papert (Seymour). Jaillissement de l’esprit : ordinateurs et apprentissage”, Revue française de pédagogie, vol. 62, no. 1, pp. 94–96, 1983. [PER 00] PERRENOUD P., “A arte de construir competências”, in GENTILE P., BENCINI R. (eds), Nova Escola, Brazil, pp. 19–31, available at: http://www.unige.ch/fapse/SSE/teachers/ perrenoud/php_main/php_2000/2000_30.html, September 2000. [PER 03] PERAYA D., “De la correspondance au campus virtuel : formation à distance et dispositifs médiatiques”, in CHARLIER B., PERAYA D. (eds), Technologie et innovation en pédagogie. Dispositifs innovants de formation pour l’enseignement supérieur, De Boeck, pp. 79–92, available at: http://archive-ouverte.unige.ch/unige:29016, 2003. [PER 04a] PERNIN J.-P., LEJEUNE A., “Dispositifs d’apprentissage instrumentés par les technologies : vers une ingénierie centrée sur les scénarios”, Technologies de l’Information et de la Connaissance dans l’Enseignement Supérieur et de l’Industrie, University of Technology of Compiègne, pp. 407–414, available at: https://halshs .archives-ouvertes.fr/edutice-00000730/document, October 2004. [PER 04b] PERNIN J.-P., LEJEUNE A., “Nouveaux dispositifs instrumentés et mutations du métier de l’enseignant”, 7e biennale de l’Education, Lyon, available at: http://www.inrp.fr /Acces/Biennale/7biennale/Contrib/longue/7134.pdf, April 2004. [PER 04c] PERRENOUD P., “Qu’est-ce qu’apprendre ?”, Enfance & Psy, no. 24, pp. 9–17, available at: http://www.unige.ch/fapse/SSE/teachers/perrenoud/php_main/php_2004/ 2004_08.html, 2004. [PER 05] PERAYA D., “La formation à distance : un dispositif de formation et de communication médiatisées. Une approche des processus de médiatisation et de médiation”, TDR, no. 0a, p. 12, available at: http://www.revue-tice.info/document .php?id=520, 2005. [PER 08] PERAYA D., CHARLIER B., DESCHRYVER N., “Dispositifs hybrides”, in PERNIN J.-P., GODINET H. (eds), ACI “Terrains, Techniques, Théories”, Projet ACTEURS (Activités collectives et tutorat dans l’enseignement universitaire : réalités, scénarios et usages des TICE), INRP, Lyon, pp. 14–24, available at: http://eductice.ens-lyon.fr/EducTice /recherche/archives/acteurs/RapportProjetACTEURS.pdf, 2008. [PER 14] PERAYA D., BONFILS P., “Détournements d’usages et nouvelles pratiques numériques : l’expérience des étudiants d’Ingémédia à l’université de Toulon”, Sticef, no. 21, available at: http://sticef.univ-lemans.fr/num/vol2014/19-peraya-epa/sticef_2014 _NS_peraya_19.htm#fn4, 2014. [PIA 69] PIAGET J., Psychologie et pédagogie, Denoël, Paris, 1969.

References

283

[PIA 99] PIASER A., Représentations professionnelles à l’école, Particularités selon le statut : enseignant, inspecteur, Doctoral thesis, University of Toulouse-Le Mirail, Toulouse, 1999. [PIA 04] PIAGET J., INHELDER B., La psychologie de l’enfant, PUF, Paris, 2004. [POC 71] POCZTAR J., Théories et pratique de l’enseignement programmé : Guide des enseignants, Unesco, Paris, 1971. [POP 94] POPPER K., Toute vie est résolution de problèmes, Actes Sud, Paris, 1994. [POR 10] PORTELLI P., “Enseignement numérique : du sur mesure”, Savoir(s), vol. 8, no. 10, available at: http://www.unistra.fr/uploads/media/SAVOIR_S__8.pdf, 2010. [POT 03] POTHIER M., Multimédias, dispositifs d’apprentissage et acquisition des langues, Ophrys, Paris, 2003. [POY 12] POYET F., GENEVOIS S., “Vers un modèle compréhensif de la généralisation des usages des ENT dans l’enseignement secondaire”, Revue française de pédagogie, no. 181, pp. 83–98, available at: http://rfp.revues.org/3927, 2012. [PRO 07] PROVITOLO D., “A proposition for a classification of the catastrophe systems based on complexity criteria”, Proceedings of 4th European Conference on Complex Systems (ECCS’07), EPNACS’2007 Emergent Properties in Natural and Artificial Complex Systems, Dresden, 2007. [PRO 08] PROVITOLO D., “Théorie de l’auto-organisation critique”, Hypergeo, available at: http://www.hypergeo.eu/spip.php?article426#, 2008. [RAB 95] RABARDEL P., Les hommes et les technologies. Approches cognitives des instruments contemporains, A. Colin, Paris, 1995. [RAB 03] RABARDEL P., BOURMAUD G., “From computer to instrument system: a developmental perspective”, Interacting with Computers, vol. 15, no. 5, pp. 665–691, available at: http://iwc.oxfordjournals.org/content/15/5/665.abstract, 2003. [RAY 97] RAYNAL F., RIEUNIER A., POSTIC M., Pédagogie : Dictionnaire des concepts clés : Apprentissages, formation, psychologie cognitive, Éditions ESF, Paris, 1997. [RIC 96] RICHEY R.C., NELSON W.A., “Developmental research”, in JONASSEN D.H. (ed.), Handbook of Research for Educational Communications and Technology, Macmillan, New York, pp. 1213–1245, 1996. [ROB 12] ROBERGE A., L’apprentissage inverse : avancée ou régression ?, available at: http://cursus.edu/dossiersarticles/articles/18434/apprentissageinverseavanceeregression/, 2012. [ROC 98] ROCHET C., “Le diagramme des affinités (Méthodes KJ)”, available at: http://claude.rochet.pagesperso-orange.fr/kj/KJ.html, March 1998. [ROC 13a] ROCHET C., “Chaos et émergence”, available at: http://claude-rochet.fr/wp2/ category/definition/modelisation-des-systemes-complexes/, November 2013.

284

Modeling of Next Generation Digital Learning Environments

[ROC 13b] ROCHET C., “Comment un système deviant complexe ?”, available at: http://clauderochet.fr/wp2/sample-page/comment-un-systeme-devient-complexe/, November 2013. [ROG 95] ROGERS Y., SCAIFE M., “External cognition as an alternative framework for evaluating single, multi and virtual media”, in LEE J. (ed.), First International Workshop on Intelligence and Multimodality in Multimedia Interfaces : Research and Applications, University of Edinburgh, Edinburgh, 1995. [ROU 03] ROUX J.-P., “Socio-constructivisme et apprentissages scolaires”, in AMIGUES R. (ed.), Petit vocabulaire raisonné à l’usage des professeurs débutants, available at: http://recherche.aix-mrs.iufm.fr/publ/voc/n1/roux/index.html, 2003. [RUM 95] RUMBAUGH J., BLAHA M., PREMERLANI W. et al., Modélisation et conception orientées objet, Masson, Paris, 1995. [RUS 01] RUSSELL T., The No Significant Difference Phenomenon: A Comparative Research Annotated Bibliography on Technology for Distance Education, IDECC, Chicago, IL, available at: http://www.nosignificantdifference.org/, 2001. [SAN 04] SANTOS O.C., BOTICARIO J.G., BARRERA C., “Authoring a collaborative task extentending the IMS LD to be performed in a standard-based Adaptive learning management system called ALFANET”, Communication présentée à l’International Conference on Web Engineering, Workshop AHCW’04, Munich, Germany, available at: http://www.ii.uam.es/%7Ercarro/AHCW04/Santos.pdf, 26–30 July 2004. [SCA 89] SCARDIGLI V., “Nouvelles technologies : l’imaginaire du progrès”, in GRAS A., POIROT-DELPECH S. (eds), L’imaginaire des techniques de pointe. Au doigt et à l‘œil, L’Harmattan, Paris, pp. 97–114, 1989. [SCH 85] SCHNEUWLY B., BRONCKART J.-P., Vygotsky aujourd’hui, Delachaux & Niestlé, Paris, 1985. [SCH 95] SCHIFFMAN S.S., “Instructional systems design, five views of the field”, in ANGLIN G.J. (ed.), Instructional Technology, Past, Present, and a Future, Libraries Unlimited, Englewood, pp. 131–143, 1995. [SCR 67] SCRIVEN M., “The methodology of evaluation”, in TYLER R., GAGNE R., SCRIVEN M. (eds), Perspective of Curriculum Evaluation, Rand McNally, Skokie, pp. 39–83, 1967. [SEL 12] SELWYN N., “Making sense of young people, education and digital technology: the role of sociological theory”, Oxford Review of Education, vol. 38, no. 1, pp. 81–96, 2012. [SEN 93] SENACH B., “L’évaluation ergonomique des interfaces homme-machine : une revue de la littérature”, in SPERANDIO J.C. (ed.), L’ergonomie dans la conception des projets informatiques, Octares éditions, Toulouse, pp. 69–122, 1993. [SER 12] SERRES M., Petite Poucette, Le Pommier, Paris, 2012. [SER 15] SERRES M., Le Monde.fr – MSciences, video, available at: http://www.lemonde.fr /sciences/video/2015/01/24/michel-serres-l-ordinateur-et-la-tete-de-saint-denis_4562802 _1650684.html, 24 January 2015.

References

285

[SHA 48] SHANNON C.E., “A mathematical theory of communication”, Bell System Technical Journal, vol. 27, no. 3, pp. 379–423, 1948. [SHA 75] SHANNON C.E., WEAVER W., Théorie mathématique de la communication, Retz, Paris, 1975. [SHA 07] SHARPLES M., TAYLOR J., VAVOULA G., “A theory of learning for the mobile age”, in ANDREWS R., HAYTHORNTHWAITE C. (eds), The Sage Handbook of E-learning Research, Sage, London, pp. 221–247, 2007. [SIE 05] SIEMENS G., “Connectivism: a learning theory for the digital age”, International Journal of Instructional Technology and Distance Learning, vol. 2, no. 1, available at: http://www.itdl.org/Journal/Jan_05/article01.htm, 2005. [SIE 11] SIEMENS G., DOWNES S., “The MOOC guide”, Wikipédia, available at: https://sites.google.com/site/themoocguide/3-cck08—the-distributed-course, 2011. [SIM 91] SIMON H.A., Sciences des systèmes, Sciences de l’artificiel, Dunod, Paris, 1991. [SIM 10] SIMONIAN S., “Rencontre avec Stéphane Simonian à propos de son ouvrage. Hypertexte et processus cognitifs, quels enjeux pour l’apprentissage”, Questions Vives, vol. 4, no. 13, pp. 401–409, available at: https://questionsvives.revues.org/339, 2010. [SKI 95] SKINNER B.F., La révolution scientifique de l’enseignement, Mardaga, Brussels, 1995. [SNO 02] SNOWDEN D., “Complex acts of knowing: paradox and descriptive self-awareness”, Journal of Knowledge Management, vol. 6, no. 2, pp. 100–111, 2002. [TAR 79] TARDIEU H., NANCI D., PASCOT D., Conception d’un système d’information, Construction de bases de données, Éditions d’Organisation, Paris, 1979. [TAR 86] TARDIEU H., ROCHFELD A., COLLETTI R., La Méthode Merise, Principes et outils, Éditions d’Organisation, Paris, 1986. [TAT 03] TATTERSALL C., VOGTEN H., BROUNS F. et al., Delivering Courses Modelled Using IMS Learning Design, Open University of the Netherlands, Heerlen, available at: http://hdl.handle.net/1820/35 (consulted on 10 January 2005), 2003. [TAV 99] TAVRIS C., WADE C., Introduction à la psychologie – Les grandes perspectives, Erpi, Saint-Laurent, 1999. [TCH 09] TCHOUNIKINE P., Précis de recherche en ingénierie des EIAH, available at: http://lig-membres.imag.fr/tchounikine/Precis.html (consulted on 24 May 2015), 2009. [THO 18] THOBOIS-JACOB L., CHRISTOFFEL E., MARQUET P., “L’adhésion des étudiants à la classe inversée : une approche par le style d’apprentissage”, Sticef, 2018. [TOL 97] TOLILA P., “Quand la prospective rêve de science”, Le Banquet, la revue du (CERAP), Centre d’étude et de réflexion pour l’action politique, available at: http://www.revue-lebanquet.com/ (consulted on 12 August 2015), January 1997.

286

Modeling of Next Generation Digital Learning Environments

[TRE 05] TRESTINI M., Conception et étude d’un dispositif de formation à distance médiatisé par les technologies de l’information et de la communication dans l’enseignement supérieur : cas de la formation initiale des enseignants, Doctoral thesis, Louis Pasteur University, Strasbourg, available at: http://www.inrp.fr/vst/Recherches/DetailThese .php?these=580, 2005. [TRE 06] TRESTINI M., “Espace virtuel… et vrais apprentissages”, Cahiers pédagogiques, no. 446, pp. 51–52, 2006. [TRE 08] TRESTINI M., “Préface”, in LEMIRE G. (ed.), Modélisation et construction des mondes de connaissances : Aspects constructiviste, socioconstructiviste, cognitiviste et systémique, pp. 1–8, Presses de l’université Laval, Quebec, available at: http://www.pulaval.com/catalogue/modelisation-construction-des-mondes-connaissancesaspects-9187.html, 2008. [TRE 09a] TRESTINI M., “Étude des usages d’un cahier de textes en ligne dans l’environnement PLACE”, Symposium International : “Distances et savoirs à sept ans !”, Cned-Eifad 2009, Site du Futuroscope, Poitiers, available at: http://www.cned.fr/media/68580/symposium-marctrestini.pdf, 11 December 2009. [TRE 09b] TRESTINI M., “Préface”, in LEMIRE G. (ed.), Modélisation et construction des mondes de connaissances : Aspects constructiviste, socioconstructiviste, cognitiviste et systémique, Cahier des Lectures MCX (Modélisation de la Complexité), available at: http://www.intelligence-complexite.org/fr/cahier-des-lectures/recherche-dune-note-delecture.html?tx_mcxapc_pi1%5Baction%5D=noteDetail&tx_mcxapc_pi1%5BidNote%5 D=711&cHash=5562b2fabb662770b1ff919c52b86326, January 2009. [TRE 10a] TRESTINI M., LEMIRE G., “Environnements Constructivistes d’Apprentissage : quand les activités collectives et individuelles permettent aux apprenants de construire leurs connaissances”, Education Canada, vol. 50, no. 1, pp. 56–60, available at: https://www.edcan.ca/wp-content/uploads/EdCan-2010-v50-n1-Trestini.pdf, 2010. [TRE 10b] TRESTINI M., CHRISTOFFEL E., ROSSINI I., “Description et usages d’un nouveau dispositif de vidéo-cours : l’encre numérique”, Questions vives, Recherches en éducation, vol. 7, no. 14, pp. 119–133, available at: https://questionsvives.revues.org/656, 2010. [TRE 12a] TRESTINI M., “Causes de non-usage des TICE à l’université : des changements ? Cas des enseignants du supérieur”, Recherches & Éducations, no. 6, pp. 15–33, available at: http://rechercheseducations.revues.org/935 (consulted on 26 October 2015), 2012. [TRE 12b] TRESTINI M., COULIBALY B., ROSSINI I. et al., Évaluation du dispositif d’enseignement à distance à l’université de Strasbourg, Report, Strasbourg, France, available at: http://ife.ens-lyon.fr/vst/Rapports/DetailRapport.php?parent=actu&id=1738, 2012. [TRE 14] TRESTINI M., COULIBALY B., “Vers une industrialisation de la formation à distance à l’université”, Distances et médiations des savoirs, no. 6, available at: http://dms.revues.org/687 (consulted on 04 July 2014), 2014.

References

287

[TRE 15a] TRESTINI M., ROSSINI I., “Les MOOC : perception des acteurs français de l’enseignement en ligne”, Revue internationale des technologies en pédagogie universitaire, vol. 12, nos 1–2, pp. 93–106, available at: http://www.ritpu.ca /IMG/pdf/RITPU_v1_n01-02_93.pdf (consulted on 22 December 2015), 2015. [TRE 15b] TRESTINI M., ROSSINI I., “MOOC in the paradigm of systemic modelling of complexity: some emerging properties”, The Online Journal of Distance Education and e-Learning, vol. 4, no. 1, pp. 44–49, available at: http://www.tojdel.net/journals /tojdel/articles/v04i01/v04i01_04.pdf, 2015. [TRE 16a] TRESTINI M., COULIBALY B., ROSSINI I. et al., Appropriation sociale des MOOC en France, ISTE Editions, London, 2016. [TRE 16b] TRESTINI M., GOSSIN P., Les espaces, in Articulation et complémentarité des équipes pédagogiques et des services de documentation au cœur de la rénovation pédagogique, pp. 31–47, Report, LISEC, Strasbourg, available at: http://www.enseignementsup-recherche.gouv.fr/pid29939-cid110776/de-la-pedagogie-ala-documentation.html, 2016. [TRE 17] TRESTINI M., CABASSUT R., “Spécificités et généricités des difficultés et besoins d’aide exprimés par les inscrits à un MOOC”, Distances et médiations des savoirs, vol. 19, available at: http://dms.revues.org/1966, 2017. [TRO 05] TROUCHE L., “Construction et conduite des instruments dans les apprentissages mathématiques : nécessité des orchestrations”, Recherches en Didactique des Mathématiques, vol. 25, no. 1, pp. 91–138, 2005. [TUK 77] TUKEY J., Exploratory Data Analysis, Addison-Wesley, Reading, 1977. [VAL 75] VALERY P., “Cahiers”, Encyclopédie de la Pléiade, Gallimard, Paris, 1975. [VAR 79] VARELA F., “Principles of biological autonomy”, in VERALA F. (ed.), Autonomie et connaissance. Essai sur le vivant, Le Seuil, Paris, 1979. [VER 00] VERGNAUD G., Lev Vygotski : Pédagogue et penseur de notre temps, Hachette, Paris, 2000. [VIA 13] VIALLON P., “La communication touristique, une triple invention”, Mondes du Tourisme, no. 7, pp. 2–11, available at: http://tourisme.revues.org/171, 2013. [VIL 07] VILLIOT-LECLERCQ E., Modèle de soutien à l’élaboration et à la réutilisation de scénarios pédagogiques, Doctoral thesis, Montreal University, Canada, 2007. [VIR 07] VIRKKUNEN J., “Le développement collaboratif d’un nouveau concept pour une activité”, @ctivités, vol. 4, no. 2, pp. 151–157, available at: http://activites.revues.org/1758, 2007. [VON 56] VON BERTALANFFY L., General System Theory, General Systems Yearbook 1, Brazziler, New York, 1956. [VON 60] VON FOERSTER H., “On self organizing systems and their environments”, in YOVITS M., CAMERON S. (eds), Self-Organizing Systems, Pergamon Press, London, pp. 31–50, 1960.

288

Modeling of Next Generation Digital Learning Environments

[VON 66] VON NEUMANN J., “Theorie of self-reproducing automata”, in BURKS A.W. (ed.), University of Illinois Press, Urbana, 1966. [VON 12] VON BERTALANFFY L., Théorie générale des systèmes (translation by CHABROL J.B.), Dunod, Paris, 2012. [VYG 78] VYGOTSKY L., Mind in Society: the Development of Higher Psychological Processes, Harvard University Press, Cambridge, 1978. [VYG 85] VYGOTSKY L., “La méthode instrumentale en psychologie”, in BRONCKART J.-P., SCHNEUWLY B. (eds), Vygotski aujourd’hui, Delachaux & Niestlé, Paris, 1985. [VYG 97] VYGOTSKY L., Pensée et langage (translation by SEVE F.), La Dispute, Paris, 1997. [WAL 09] WALLISER B., Emergent Phenomena, Document, ENPC-EHESS, Paris, available at: http://www.pse.ens.fr/users/walliser/, 2009. [WAL 12] WALDVOGEL F.A., BALAVOINE J.-F., PERONE N. et al., “Les malades complexes : de la théorie des systèmes complexes à une prise en charge holistique et intégrée”, Revue Médicale Suisse, no. 340, pp. 1022–1024, available at: http://www.revmed.ch/rms/2012 /RMS-340/Les-malades-complexes-de-la-theorie-des-systemes-complexes-a-une-prise-en -charge-holistique-et-integree#rb3, 2012. [WAT 72a] WATSON J., Le béhaviorisme, CEPI, Paris, 1972. [WAT 72b] WATZLAWICK P., HELMICK BEAVIN J., JACKSON D., Une logique de la communication, Le Seuil, Paris, 1972. [WIE 48] WIENER N., Cybernetics, Hermann et Cie, Paris, 1948. [WIL 00] WILEY D.A., “Connecting learning objects to instructional design theory: a definition, a metaphor, and a taxonomy”, in WILEY D.A. (ed.), The Instructional Use of Learning Objects (Association for Educational communications and Technology), Paper, Utah State University, Logan, 2000. [ZIN 03] ZIN J., “La complexité et son idéologie”, available at: https://jeanzin.fr/ecorevo /sciences/complexi.htm (consulted on 26 August 2016), 1 May 2003. [ZNI 12] ZNIBER N., Modèle orienté service pour la conception de parcours pédagogiques personnalisés, Doctoral thesis, University of Aix-Marseille, available at: https://tel.archives-ouvertes.fr/file/index/docid/660281/filename/these-najlaa_zniber.pdf, 2012.

Index

A, B, C activity collaborative, 121, 127 instrumented, 13, 37, 54, 55, 58, 60 theory, 68, 72, 103, 106 analysis of activity, 58 approach content-centered, 83 process-centered, 83 black box, 182, 184–186, 188, 197 cognitive psychology, 12 computer-assisted learning, 2, 12, 15, 18, 19 connectivism, 27, 28, 30–32 constructivist learning environment (CLE), 120, 122, 123

DWS in schools, 123, 125, 133 EIAH, 54 eight functions, 228, 231 emerging phenomenon, 28, 251 EML language, 66, 85 expansion of the model, 114, 116, 120, 123, 125 exploratory analysis of data, 224, 225 forms of mediation, 103 H, I, L homeostatic nature, 251 language modeling, 208 UML, 190, 194, 197, 206, 209, 239 level of understanding, 243, 246–250 long-distance learning, 116

D, E, F descriptive statistics, 224, 225 diagram activity, 128, 129 class, 131 data flow, 202, 203, 205 package, 228 sequence, 240 state transition, 239 use case, 228

M, N, O mediation collective, 103 individual, 103 microworlds, 12, 19, 21 MISA, 134 model adaptive, 81, 100, 101 analysis, 67, 72, 81

Modeling of Next Generation Digital Learning Environments: Complex Systems Theory, First Edition. Marc Trestini. © ISTE Ltd 2018. Published by ISTE Ltd and John Wiley & Sons, Inc.

290

Modeling of Next Generation Digital Learning Environments

archetype model, 183, 189, 197, 198, 201, 205, 207, 211 design, 66, 71, 72, 97, 106 learning engineering, 81, 96 object, 65, 80 pedagogical engineering, 81, 83 theoretical, 65 network, 2, 27, 29, 30, 32 object-oriented modeling, 197, 202, 206–210, 239 open system, 40–42, 50–53 oscillations, 249, 250

P, R, S, T procedure cybernetic, 181 structuralist, 181 programmed instruction, 17, 18 representations professional, 131 social, 131 resistance to models, 126, 132 script, 239, 243, 245 systemic paradigm, 39 Tablet-PC, 112–115 technopedagogical innovations, 109 tracking tools, 209, 212–214, 224

Other titles from

in Science, Society and New Technologies

2018 BARTHES Angela, CHAMPOLLION Pierre, ALPE Yves Evolutions of the Complex Relationship Between Education and Territories (Education Set - Volume 1) BÉRANGER Jérôme The Algorithmic Code of Ethics: Ethics at the Bedside of the Digital Revolution (Technological Prospects and Social Applications Set – Volume 2) DUGUÉ Bernard Time, Emergences and Communications (Engineering, Energy and Architecture Set – Volume 4) GEORGANTOPOULOU Christina G., GEORGANTOPOULOS George A. Fluid Mechanics in Channel, Pipe and Aerodynamic Design Geometries 1 (Engineering, Energy and Architecture Set – Volume 2) GEORGANTOPOULOU Christina G., GEORGANTOPOULOS George A. Fluid Mechanics in Channel, Pipe and Aerodynamic Design Geometries 2 (Engineering, Energy and Architecture Set – Volume 3)

LELEU-MERVIEL Sylvie Informational Tracking (Traces Set – Volume 1) SALGUES Bruno Society 5.0: Industry of the Future, Technologies, Methods and Tools (Technological Prospects and Social Applications Set – Volume 1)

2017 ANICHINI Giulia, CARRARO Flavia, GESLIN Philippe, GUILLE-ESCURET Georges Technicity vs Scientificity – Complementarities and Rivalries (Social Interdisciplinarity Set – Volume 2) DUGUÉ Bernard Information and the World Stage – From Philosophy to Science, the World of Forms and Communications (Engineering, Energy and Architecture Set – Volume 1) GESLIN Philippe Inside Anthropotechnology – User and Culture Centered Experience (Social Interdisciplinarity Set – Volume 1) GORIA Stéphane Methods and Tools for Creative Competitive Intelligence KEMBELLEC Gérald, BROUDOUS EVELYNE Reading and Writing Knowledge in Scientific Communities– Digital Humanities and Knowledge Construction MAESSCHALCK Marc Reflexive Governance for Research and Innovative Knowledge (Responsible Research and Innovation Set - Volume 6) PARK Sejin, GUILLE-ESCURET Georges Sociobiology vs Socioecology – Consequences of an Unraveling Debate (Interdisciplinarity between Biological Sciences and Social Sciences Set Volume 1)

PELLÉ Sophie Business, Innovation and Responsibility (Responsible Research and Innovation Set - Volume 7)

2016 BRONNER Gérald Belief and Misbelief Asymmetry on the Internet EL FALLAH SEGHROUCHNI Amal, ISHIKAWA Fuyuki, HÉRAULT Laurent, TOKUDA Hideyuki Enablers for Smart Cities GIANNI Robert Responsibility and Freedom (Responsible Research and Innovation Set - Volume 2) GRUNWALD Armin The Hermeneutic Side of Responsible Research and Innovation (Responsible Research and Innovation Set - Volume 5) LAGRANA Fernando E-mail and Behavioral Changes – Uses and Misuses of Electronic Communications LENOIR Virgil Cristian Ethical Efficiency – Responsibility and Contingency (Responsible Research and Innovation Set - Volume 1) MAESSCHALCK Marc Reflexive Governance for Research and Innovative Knowledge (Responsible Research and Innovation Set - Volume 6) PELLÉ Sophie, REBER Bernard From Ethical Review to Responsible Research and Innovation (Responsible Research and Innovation Set - Volume 3)

REBER Bernard Precautionary Principle, Pluralism and Deliberation – Sciences and Ethics (Responsible Research and Innovation Set - Volume 4) VENTRE Daniel Information Warfare – 2nd edition