Learning teaching in teaching: Online reinforcement learning for intelligent tutoring

Abstract

Intelligent tutoring has started to, and will play an important role in education and training. A challenging task in building an intelligent tutoring system (ITS) is to create and maintain an optimal teaching strategy. In this paper, we present a new technique for addressing this challenge. We cast an intelligent tutoring system as a Markov decision process (MDP), and apply a reinforcement learning (RL) algorithm to learn the optimal teaching strategy through interactions between the system and students. This technique enables the system to teach a student based on his/her studying states, and allows the system to learn the optimal teaching strategy in an online fashion. © 2014 Springer-Verlag.