Learning Pedagogical Policies from Few Training Data

Learning a pedagogical policy in an Adaptive Educational System (AIES) fits as a Reinforcement Learning (RL) problem. However, to learn pedagogical policies requires to acquire a huge amount of experience interacting with the students, so applying RL to the AIES from scratch is infeasible. In this paper we describe RLATES, an AIES that uses RL to learn an accurate pedagogical policy to teach a course of Data Base Design. To reduce the experience required to learn the pedagogical policy, we propose to use an initial value function learned with simulated students, whose model is provided by an expert as a Markov Decision Process. Empirical results demonstrate that the value function learned with the simulated students and transferred to the AIES is a very accurate initial pedagogical policy. The evaluation is based on the interaction of more than 70 Computer Science undergraduate students, and demonstrates that an efficient guide through the contents of the educational system is obtained.