Um modelo de ambiente interativo de aprendizagem baseado numa arquitetura multiagentes.

Abstract

Most of the research in the design of Artificial Intelligence in Education (AI-ED) system involves tutoring interactions between two main entities: a computer tutor and a human learner. In this context, one of the central and fundamental issues is the adaptability of the computer tutor to the learner's individual needs (as perceived by the tutor), that is, the possibility that the tutor provides the learner with appropriate and individualized actions in teaching/learning situations. This adaptability has been widely assumed as a major requirement to achieve an effective learning process. The present research is focused in this direction, mainly aiming the design of a model of a multi-agent computer-based interactive learning environment. In particular, this environment model aims to support pedagogical interactions between a human learner and a society of artificial tutoring agents, and with this society and a human expert society. We make considerations about the domain knowledge structure and defining a multidimensional view on this domain. Then, we adopt a multi-agent approach based on this view and defining a society of tutoring agents. So, we develop the main elements to support the interactions involved. These elements are related to agent model, agent architecture, languages and protocols to support interactions among agents. Finally, it is important to claim that the designed learning environment and its multiagent architecture mean a horizontal contribution in this thesis. A vertical contribution of the present research is related to proposition of a multidimensional model of domain knowledge, of a multi-agent system, and of distributed model to support diagnosis of learner's actions.