International audience

A lack of access to learning opportunities is among the main reasons for social exclusion of people with cognitive disabilities in the professional field. It has been accentuated in the last few years by the rapid development of online learning strategies that limit the physical presence of students and instructors. Personalized systems can help overcome access barriers in learning for people with disabilities by helping them learn autonomously without the direct intervention of tutors. However, despite the wide range of research initiatives in this area to deal with individual differences, learners with cognitive disabilities still suffer from a lack of consideration of their conditions. Indeed, most e-learning platforms have been designed without taking into account users with multiple profiles and impairments. This paper systematically reviewed papers to provide insights about personalization within e-learning systems. We highlight personalization goals and approaches and review whether they improve learning accessibility and outcomes. The objective is to discuss the strengths and weaknesses of current research and suggest some opportunities to improve research in personalized e-learning systems for people with cognitive disabilities.

International audience

International audience

International audience

International audience

Identify the serious games that best meet the needs and expectations of teachers and pedagogical objectives of their courses remains a necessity about the integration of serious games in the learning process. Indeed, several serious games have developed in recent years, and it is often difficult for a teacher, not a computer scientist in particular, to find and choose a game that meets its specific needs. Our aim is to develop models and tools to support teachers/trainers in their choice of serious games through knowledge extraction of educational objectives, considering user feedback and their traces of interaction with the game.

International audience

In this article, we propose a set of models and a prototype to capitalize and share knowledge of expert tutors using training simulators. This is a particularly important issue in contexts with strong stakes such as training of operators of nuclear power plants, where operators’ accreditation strongly depends on skills of tutors and tolerate no errors. In such a context, observation, analysis and debriefing of interactions of trainees’ operators are complex activities especially for the young tutors who do not have the expertise of confirmed tutors. Based on digital traces, our approach consists in providing a visual synthesis of trainees’ activities by using the knowledge of experts. Such a synthesis, showing the relationships between low-level traces and high-level behaviors, enables tutors to enhance their understanding and better analyse the activity in order to prepare the debriefing. Our approach has been implemented in a prototype, called D3KODE, which was evaluated according to a comparative protocol conducted with a team of tutors from EDF Group (Electricity Of France). The result demonstrated that the visual synthesis and the higher informations provided by D3KODE helped the intructors to confirm/validate more easily realizations and no-realizations of educational objectives trainees and facilitated the exchanges between tutors and trainees.

Adaptive and personalized behavior is becoming essential and desirable in Human-Robot Interactive systems. We are interested in adaptive robots that learn from interaction traces (previous interactions with users). Our proposal is based on types of interactions where users express their level of satisfaction through feedback. Indeed, depending on the situation of interaction and the user himself, the robot behavior should adjust, and therefore can be judged, differently. From interaction traces (including robot actions and users' feedback), we aim to extract adaptation rules that give the dependencies between certain attributes of the interaction situation and/or the user profile, and the level of user satisfaction. We propose two learning algorithms to learn these adaptation rules. The first algorithm is direct, certain and optimal but slow to converge. The second is able to detect the importance of certain attributes in the adaptation process. It generalizes adaptation rules on unknown situations and to first time users, which makes it an approach with risk. We detail in this paper, our proposed model, both learning algorithms, and an evaluation of the learned rules from both algorithms by simulations and through a scenario with real users.

International audience

International audience

Analysing learners' behaviour continuously and under ecological conditions can help designers, trainers and teachers to analyse, design, validate, and also to adapt and personalize the learning game. Metrics methods propose to collect any interactions between a user and the game. While classical metrics methods fall within quantitative approaches, we aim to extract some qualitative information on high-level behaviours. This paper is focused on learners' engaged-behaviours. Thus, to identify and to qualify learners' engagement from their traces of interaction, we combine a theoretical work on engagement and engaged-behaviours, the Self-Determination Theory, the Activity Theory and a trace framework. We implemented this approach on 12 players' interaction data collected during four months. As a result, we identified and qualified four activities that refer to different types of engaged-behaviours. Thus, this user study show the feasibility and the validity of the proposed approach.

The task of generating personalized learning scenarios, in an automatic fashion, for learners is a complicated task. The difficulty is compounded when these scenarios are to be integrated within a serious game dedicated to people with disabilities. In this case, the specificities of the learner's preferences and competencies should also be taken into account in addition to the learning domain and the serious game specificities. Building on our previous work, we present in this paper the platform GOALS. This platform allows educational designers to generate learning scenarios for learners, manage learning domain and serious game knowledge, as well as, the learners' profiles. We also present two evaluations in the context of project CLES (Cognitive Linguistic Elements Stimulations). The objectives of these evaluations are to verify the functionality of GOALS, to verify the project CLES's knowledge and to study the impact of the generated scenarios on the learning of real-world learners.

Service robots have become increasingly important subjects in our lives. However, they are still facing problems like adaptability to their users. While major work has focused on intelligent service robots, the proposed approaches were mostly user independent. Our work is part of the FUI-RoboPopuli project, which concentrates on endowing entertainment companion robots with adaptive and social behaviour. In particular, we are interested in robots that are able to learn and plan so that they adapt and personalize their behaviour according to their users. Markov Decision Processes (MDPs) are largely used for adaptive robots applications. However, one challenging point is reducing the sample complexity required to learn an MDP model, including the reward function. In this article, we present our contribution regarding the representation and the learning of the reward function through analysing interaction traces (i.e. the interaction history between the robot and their users, including users' feedback). Our approach permits to generalise the learned rewards so that when new users are introduced, the robot may quickly adapt using what it learned from previous experiences with other users. We propose, in this article, two algorithms to learn the reward function. The first is direct and certain, the robot applies with a user what it learned during interaction with same kind of users (i.e. users with similar profiles). The second algorithm generalises what it learns to be applied to all kinds of users. Through simulation, we show that the generalised algorithm converges to an optimal reward function with less than half the samples needed by the direct algorithm.

This work focuses on the sharing of experiences between learners with different profiles. This is to allow learners with different skills, abilities or preferences to exchange, among themselves, the traces of their own activities. In this context, we are particularly interested in the transformation process that adapts the shared traces according to the profile of its target user. In this article, we propose models for representing trace, user profile and adaptation knowledge. In order to bring flexibility and adaptability to our approach, we propose a model of knowledge extraction from a Trace-Based Management System, which contains the activity traces of several users. The results of applying our approach to an e-learning training are also presented.

Our work addresses the development of a Serious Game for the diagnostic and learning of persons with cognitive disabilities. In reality, many studies have shown that young people, especially children, are attracted towards computer games. Often, they play these games with great interest and attention. Thus, the idea of using serious games to provide education is attractive for most of them. This work is situated in the context of Project CLES. This project, in collaboration with many research laboratories, aims at developing an Adaptive Serious Game to treat a variety of cognitive handicaps. In this context, this article presents a system that generates learning scenarios keeping into account the user's profile and their learning objectives. The user's profile is used to represent the cognitive abilities and the domain competences of the user. The system also records the user's activities during his/her interaction with the Serious Game and represents them in interaction traces. These traces are used as knowledge sources in the generation of learning scenarios.

Our work deals with the subject of activity observation in full scope simulators. In this paper we present a Modelled Trace approach for designing a methodological framework and associated tools to manage the observation process and the corresponding analysis activity.

This work addresses issues relevant to the project CLES (Cognitive and Linguistic Element Stimulation) which aims to develop a serious game for diagnosis and training of children with cognitive disabilities. In this context, our objective is to propose a system capable of generating adaptive scenarios taking into account the user's profile. A scenario here is a suite of pedagogical activities allowing the learner to achieve his/her pedagogical goal(s). The system is intended to be as generic as possible i.e. capable of being utilized with a variety serious games. Therefore, we've identified and separated different types of knowledge represented by the system namely, the domain concepts, the pedagogical resources and the serious game resources.

In the context of e-learning, this work focuses on sharing experiences between users (learners, tutors or designers). This article focuses particularly on the transformation process to adapt the shared traces according to the profile of their target users. The trace is defined as history of user actions collected in real time from his/her interactions with the computer environment. Thus, we propose models for representing trace, user profile and adaptation knowledge. In order to bring flexibility and adaptability to our approach, we propose a model of knowledge extraction from a Trace-Based Management System (TBMS) representing the activity traces of several users.

This work is a part of an ongoing project that focuses on potential applications of an interactive system that helps children with autism. Autism is classified as a neurodevelopmental disorder that manifests itself in markedly abnormal social interaction, communication ability, patterns of interests, and patterns of behavior [1]. Children with autism are socially impaired and usually do not attend to the people around them. An interesting point which characterized children with autism is that they are unable to choose which event is more or less important. As a consequence they are often saturated because of too many stimuli and thus they adopt an extremely repetitive, unusual, self-injurious, or aggressive behaviour. Recently, a new trend of using human computer interface (HCI) technology and computer science in the treatment of autism has emerged [2, 3]. The platform we developed helps children with autism to focus their attention on a specific task. In this article, we only present the attention analysis system which is a part of a more general system that used a multi-agent architecture [4]. Each task proposed on our system fit to each child, is reproducible and evolutive following a specific scenario defined by the expert. This scenario takes into account age, ability, and degree of autism of each child. In order to focus a child's attention onto the relevant object, our system displays or plays specific stimulus; once again the specific stimulus is defined for each child. Symbol or sound represents an emotional and satisfaction value for the child. The major problem is to define the correct moment when the system has to (dis)play this signal. We tackle this problem by defining a robust measure of attention. This measure is defined by analyzing the gaze direction and the face orientation, and incorporating the child's specific profile. Following expert directives, our system helps children to categorize elementary perception (strong, smooth, quick, slow, big, small...). Our objective is that children re-use these classifications in others situations.