A Recommender Model for the Personalized Adaptive CHUNK Learning System

Abstract

Recommender systems attempt to influence one’s behavior based on explicit and implicit information provided by the users of the system. Users who take part in e-commerce or watch cat videos online will be familiar with this concept. Different algorithms exist that determine what objects or concepts to recommend to users, but every one of them has the similar goal of providing a good recommendation. In this context, good means that the recommendation will be user relevant suggesting accurate topics, and will influence the user’s behavior. Additionally, a good recommendation system is adaptive, consistently seeking feedback from the user. Feedback is then used to make the next recommendation better. In this work, we develop a recommendation methodology for an existing personalized learning system, where both content and teaching methodology options are presented to the user. Our methodology provides solutions to both the user and the network coldstart problems, where little up-front information is available in order to make good recommendations. Using real system data, we show how our method recommends the most relevant learning topics and styles and incorporates user feedback to improve future recommendations.