Exploring the Space of Whole-Group Case Retrieval in Making Group Recommendations

Case-Based Reasoning has been studied as a methodology to support ratings-based collaborative recommendation, but this predominantly targets the context of an individual end-user. There are, however, many circumstances where several people participating together in a group activity could benefit from recommendations tailored to the group as a whole. Group recommendation has received comparatively little attention overall, and recent research has largely focused on making straightforward individual recommendations for each group member and then aggregating the results. But this examines only the context of the target group, and does not take advantage of other, previous group contexts as a first-class element of the knowledge base. Recent research investigated how case-based reasoning approaches can be applied to retrieve and reuse whole previous groups as a basis for recommendation and showed an advantage over traditional aggregation approaches. In this paper we focus on further exploration of the space. We present our approach for case-based group recommendation, as well as evaluation results across conditions for group size and homogeneity. Results show that foundational group-to-group approaches outperform individual-to-group recommendations across a wide range of group contexts.