Files
Abstract
This dissertation examines the effects of online social recommendation systems (RS) on consumer preference similarity under different social network structural properties. There is a debate on recommendation systems effects on consumer preference similarity. One view suggests that RS heterogenizes consumer preferences (i.e., making consumers less similar); while another view proposes an opposite effect (i.e., RS makes consumers to be more similar). This study first resolves the debate by revealing RSs effects depends on: (1) the level of the analysis; and (2) the type of recommendation used. Secondly, based on a large archival data set from an online music recommendation provider, we examine the effects of social recommendation systems (SRS) on consumer preference diversity and preference similarity at three different levels: individual level, ego-network level, and cluster level. Our findings show that the homogenizing and heterogenizing effects of a SRS depend on consumers social network structural properties. At the individual level, SRS tends to diversify consumer preferences as consumers centrality increases. At the ego-network level, a central consumer in a network (e.g., a consumer with many connections) is more likely to have preferences that are different to her connections. Moreover, SRSs effect on preference similarity is non-linear: it becomes weaker as a consumer has more connections. At the cluster level, SRS homogenizes consumer preferences if consumers in the cluster connect to each other densely. This dissertation provides important insights to both academics and practitioners.