Toward Optimal Pedagogical Action Patterns by Means of Partially Observable Markov Decision Process

Good pedagogical actions are key components in all learning-teaching schemes. Automate that is an important Intelligent Tutoring Systems objective. We propose apply Partially Observable Markov Decision Process (POMDP) in order to obtain automatic and optimal pedagogical recommended action patterns in benefit of human students, in the context of Intelligent Tutoring System. To achieve that goal, we need previously create an efficient POMDP solver framework with the ability to work with real world tutoring cases. At present time, there are several Web available POMDP open tool solvers, but their capacity is limited, as experiments showed in this paper exhibit. In this work, we describe and discuss several design ideas toward obtain an efficient POMDP solver, useful in our problem domain.