Stephen Downes

Absolutely fascinating paper that studies community (or cluster) formation in networks. Unlike much of the literature in the field, this study begins with a set of 'ground-truth communities' - these are "networks where nodes explicitly state their community memberships." This would be things like groups in Facebook or Orkut, for example, "focused on specific topics, interests, hobbies, affiliations, and geographical regions." This is where we would expect to find the most dense cluster of connections. But in fact, the most dense cluster of connections occurs between ground-truth communities. This leads the authors to postulate that "in a sense we are discovering pluralistic homophily where the similarity of one node to another is the number of shared affiliations, not just their similarity along a single dimension." This they describe using a Community-Affiliation Graph Model (AGM). (Cite as: arXiv:1205.6228v1 [cs.SI]) Via Computational Legal Studies.