Reflektor

Music is intrinsically linked to our social lives. As more music becomes available through streaming services, deciding what music is appropriate for social events becomes increasingly challenging and nuanced. While prior work has considered the social role of music and the creation of music playlists for user contexts, how individuals utilize music to create social contexts is an area that has largely gone unexplored. To investigate this topic, we created and evaluated a prototype music recommender system called Reflektor. Reflektor interactively visualizes users' chat conversations to generate music playlists. Our analysis of user conversations with Reflektor uncovered distinct strategies participants use to create the ambiance and conduct for social contexts. Our findings help to illuminate mismatches in the way metadata and recommendation systems align with user strategies to create social context. We elaborate on these strategies and discuss design implications for future collaborative music recommender systems.