Community Detection and Homophily in Music Sharing Online Social Networks: A Study of This Is My Jam

Abstract

Since the rise of online social networks, online activity, including sharing, discovery and consumption of a diverse range of media, is increasingly driven by online social interactions. The music industry, in particular, has been at the forefront of this transformation. With the increasing importance that online social networks play in music consumption choices, it becomes pertinent to investigate the social dynamics that govern these networks. In this study, we investigate the application of community detection to this task. In particular, we obtain the network of followers and longitudinal data on user positing and liking activities over a two month observation period of a small music sharing social network, This Is My Jam. We apply community detection, using two state-of-the-art community detection algorithms, Infomap and GANXiS, to the follower network and to an affinity network constructed based on relative liking activity frequencies amongst users. We evaluate community structure in both networks and find that only the affinity network displays a strong community structure. We then investigate community-level music interest homophily in the communities of the affinity network by performing an in-depth comparison of user similarity within communities to user similarity between communities, using a variety of similarity measures to quantify musical taste similarity based on posting activities of users. Our findings are supportive of community-level homophily within well-partitioned communities and are robust across similarity measures. Our results indicate that community-level homophily is primarily due to specific musical genre interest groups. These results serve to inform future work seeking to incorporate community-level information regarding users of music sharing online social networks.