A two-level multi-agent architecture for a distance learning environment

In this paper some results achieved in the domain of multi-agent methodology applied to educational systems are presented. We have developed a distance learning environment based on a two-level multi-agent architecture. Two-levels are distinguished since the application is composed by two multi-agent systems (MAS). The higher-level MAS is composed of cognitive agents, which provide the main functions of the system. The lower-level MAS is composed of a large number of reactive agents responsible for diagnosing students' conceptions. The two levels communicate through tutor agents, whose educational decisions are based on emergent results coming from the lower level. Besides the multi-agent approach, our work is founded on the emergent theory and employs a mechanism of voting for capturing group decision. This text presents methodological and theoretical aspects of our platform, the architecture proposed and the first prototype implemented for teaching geometry proof.