Toward a social graph recommendation algorithm: Do we trust our friends in movie recommendations?

Abstract

Social networks provide users with information about their friends, their activities, and their preferences. In this paper we study the effectiveness of movie recommendations computed from such communicated preferences. We present a set of social movie recommendation algorithms, which we implemented on top of the Facebook social network, and we compare their effectiveness in influencing user decisions. We also study the effect of showing users a justification for the recommendations, in the form of the profile pictures of the friends that caused the recommendation. We show that social movie recommendations are generally accurate. Furthermore, 80% of the users that are undecided on whether to accept a recommendation are able to reach a decision upon learning of the identities of the users behind the recommendation. However, in 27% of the cases, they decide against watching the recommended movies, showing that revealing identities can have a negative effect on recommendation acceptance. © 2012 Springer-Verlag.