Learn and Evolve the Domain Model in Intelligent Tutoring Systems - Approach Based on Interaction Traces

Majority of the systems developed in the field of education aim to tailor information presented to their users according to them as efficiently as possible. Often these systems use a User Model, a Domain Model and an Interaction Model to provide the adaptation/personalization effect. However, any deficiencies in these models directly influence the quality of adaptation the system provides. In order to address this issue, we propose to use modeled user’s interaction traces to detect these deficiencies in the domain model. Furthermore, the use of these interaction traces will help us in proposing a more correct domain model according to the user’s competence. We had tested our approach by creating an educational system. The overlay modeling approach is used with Bayesian Networks to model our domain model. The results of the conducted experiments are also presented in this article.