A Probabilistic Model for Student Knowledge Diagnosis in Learning Environments

Nowadays, the way learners acquire and improve their knowledge has changed compared to several years ago. The use of new technologies, in conjunction with Artificial Intelligence (AI) and psychological research in the field of education, has made it possible to obtain a new generation of learning environments, which are more effective than traditional methods[1]. These intelligent tutoring environments are computer assisted tools capable of both representing knowledge and directing a learning strategy. Thus they are able to mimic the behavior of an expert, teaching the students how to apply knowledge and estimating their knowledge from how they interact with the system. As a result, learners can be redirected to the most appropriate action. To achieve these goals, the maintenance and use of a model that reflects the student knowledge is vital. Such a model allows the student weaknesses and strengths to be ascertained and a pedagogical strategy for appropriate instruction to be developed. We have previous experience in the development of educational tools [2]. Nevertheless, this system is focused on assessing declarative knowledge using tests. We want to extend our previous work to cover both declarative and procedural knowledge assessment. For this reason, the main goal of our current research is to study existing approaches for building student models in problem solving environments, and to develop a novel methodology or framework to facilitate the elicitation of diagnosis modules for such environments. In this line, we have explored existing student modeling techniques and have tried to identify how we could improve them to create a versatile student model (in the context of the application domain), with accurate estimations of the student knowledge.