CurriculumTutor: An Adaptive Algorithm for Mastering a Curriculum

An important problem in an intelligent tutoring system (ITS) is that of adaptive sequencing of learning activities in a personalised manner so as to improve learning gains. In this paper, we consider intelligent tutoring in the learning by doing (LbD) setting, wherein the concepts to be learned along with their inter-dependencies are available as a curriculum graph, and a given concept is learned by performing an activity related to that concept (such as solving/answering a problem/question). For this setting, recent works have proposed algorithms based on multi-armed bandits (MAB), where activities are adaptively sequenced using the student response to those activities as a direct feedback. In this paper, we propose CurriculumTutor, a novel technique that combines a MAB algorithm and a change point detection algorithm for the problem of adaptive activity sequencing. Our algorithm improves upon prior MAB algorithms for the LbD setting by (i) providing better learning gains, and (ii) reducing hyper-parameters thereby improving personalisation. We show that our tutoring algorithm significantly outperforms prior approaches in the benchmark domain of two operand addition up to a maximum of four digits.