Mixed-Initiative Friend-List Creation

Abstract

Friend lists group contacts in a social networking site that are to be treated equally in some respect. We have developed a new approach for recommending friend lists, which can then be manually edited and merged by the user to create the final lists. Our approach finds both large networks of friends and smaller friend groups within this network by merging virtual friend cliques. We have identified new metrics for evaluating the user-effort required to process friend-list recommendations, and conducted user studies to evaluate our approach and determine if and how the recommended lists would be used. Our results show that (a) our approach identifies a large fraction of the friend lists of a user, and seeds these lists with hundreds of members, few of which are spurious, and (b) users say they would use the lists for access control, messaging, filling in friend details, and understanding the social structures to which they belong.