In this paper, we propose a multi-agent approach to building more Cooperative Intelligent Distance Learning Environments (CIDLE). We define a Confidence Agent in Intelligent Tutoring System (ITS) in such a way that an ITS would improve the quality and efficiency of its teaching. To achieve this goal, we propose a Confidence Intelligent Tutoring System (CITS) to manage negotiations within a community of on-line learners to improve CIDLE interactions among the participants. The proposed system can extract knowledge about domain knowledge and about learners behavior during a learning discussion. Therefore, it infers the behavior of learners, and adapts presentation of subject mater in order to improve their success rate in answering questions and boost their self-confidence during learning session. In addition, we discuss architectural problems of the CITS and their solutions..

The large amount of information now available on the Web can play a prominent role in building a cooperative intelligent distance learning environment. We propose a system to provide learners with useful information in a group discussion. Finding the right information at the right moment is quite a difficult task, especially when the learner's interests are continually updated during the discussion. This paper presents a context-based information agent that can observe conversations among a community of learners on the Web, interpret the learners' inputs, and then assess the current context of the session. The agent must be able to adapt its behavior autonomously to the changing context, build a new query to get updated information from the Web, and originate the search task. Then, it can filter the results, organizing, and presenting information useful to the learners in their current activities. We claim that specifying the context of a search better can significantly improve search results. An important task, therefore, is to assess the context. For this, we have developed dominant meaning space. That is a new set based measure to evaluate the closeness between queries and documents. Our experiments show that the proposed method greatly improves retrieval effectiveness, in terms of average overall accuracy as well as that in the top twenty documents. This work is the core component of a new pedagogical agent to help people learn tasks defined within greater Web-based tutoring systems.

Abstract- Similar to classical in-situ laboratories, remote laboratories are necessary in e-learning environments, especially in scientific and technical disciplines. This paper outlines our current research on this particular way of learning. Our research objectives consist in proposing a generic framework (independent of educational content) to allow, in one hand, tutors to integrate (both remote and virtual) laboratories in their LMS (Learning Management System), and in the other hand, to enable creation, distribution and exchange of pedagogical scenarios for practical works. So, authors are able to deal their scenarios through LCMS (Learning Content Management System) and to reuse scenarios from other authors (as for any classical e-learning content). Furthermore, scenarios are no more written for a specific apparatus, but for a class of them (inverted pendulum, optical bench, …). Two main topics are developed. First one consists of modelling laboratory and scenario structures regarding IMS-LD specification to separate content from containers. Second one consists of describing system components and functionalities using ontologies (OWL standard in our case). A prototype for automation discipline is presented.

Although, the growing number of Internet users who can play a prominent role in building collaborative groupware systems, the ability to find helpers (tutors or other learners) is still a challenging and important problem. Helpers could have a lot of useful information about courses to be taught, however, many learners fail to understand their presentations. We suggest a new filtering framework, called a pyramid collaborative filtering model, to trickle the number of helpers down to just one. The proposed pyramid has four levels. Moving from one level to another depends on three filtering techniques: domain model filtering; user model filtering; and credibility model filtering. Our experiments show that this method greatly improves filtering effectiveness. 1

Abstract. Web-based applications with very diverse learners fail because they fail to satisfy various needs. Some people use collaborative filtering methods to analyze learners ’ profiles and provide recommendation to a new learners, but this methods provides false recommendations from beginners. We present a new method, which provides recommendations that depend on the credibility rather than the number of learners. We have designed, implemented, and tested what we call the Intelligent E-Course Agent (IECA). Our evaluation experiment shows that our approach greatly improves learners ’ knowledge and therefore presents a course that is more closely related to their needs.