Prof. Kobi Gal

One important function of e-learning systems is to sequence learning material for students. E-learning systems use data, such as demographics, past performance, preferences, skillset, etc. to construct an accurate model of each student so that the sequencing of educational content can be personalized. Some of these student features are“shallow”traits which seldom change (e.g. age, race, gender) while others are “deep” traits that are more volatile (e.g. performance, goals, in terests). In this work, we explore how reasoning about this diversity of student features can enhance the sequencing of educational content in an e-learning environment. By modeling the sequencing process as a Reinforcement Learning (RL) problem, we introduce Diversity Aware Bandit for Sequencing Educational Content (DABSEC), a novel contextual multi-armed bandit algorithm that leverages the dynamics within user features to cluster similar users together when making sequencing recommendations.