Recommendation of educational resources to groups: A game-theoretic approach
Zacharoula K. Papamitsiou, Anastasios A. Economides · 2018
In collaborative learning contexts, it is necessary to recommend educational resources to groups of students instead of individuals. However, this task is not trivial, because students in a group may not be fulfilled by the same items, yet wish to meet their own expectations. Existing approaches either merge individual profiles and recommend items accordingly, or fuse the lists of individual recommendations. Both perspectives achieve low quality performance and goodness of recommendation for majority of students in heterogeneous groups. This paper follows a game-theoretic approach for solving conflict of interest among students and recommending resources to both homogeneous and heterogeneous groups in collaborative learning contexts. The group members are the players, the resources comprise the set of possible actions, and selecting those items that will maximize all students' satisfaction - both individually and as a whole - is a problem of finding the Nash Equilibrium. During the empirical evaluation of the suggested approach compared to other state-of-the-art methods in a real dataset, the relevance of each item to its corresponding students was explored from two perspectives: the group's (as a whole) and the individual student's (within the group). Results indicate a statistically significant improvement in accuracy of predicted group and individual satisfaction, as well as in the goodness of the ranked list of recommendations.