Mining influence in recommender systems

The influence of an entity can be defined as its ability to affect the conduct, behavior, or actions of other entities. We provide evidence in this thesis that influence is important in recommender systems. Recommender systems help people find the things they care about from an unmanageably large number of choices by mining relationships between like-minded people. Discovering influential items and users can enhance the ability of a recommender system to deliver quality recommendations in various ways, including guiding a new member through the right set of items to evaluate so that the system learns her preferences effectively, and selecting reliable users for early evaluations of new items. How, then, may we discover the most influential items and users in the system? We explore several sources of insight for influence algorithms in recommender systems: social network theory, information theory, and mathematical analysis of the recommender algorithms themselves. Broadly speaking, this thesis explores the following: (a) the nature of influence in recommender systems, (b) prior research on influence in other domains and the viability of applying that research to the recommender systems domain, (c) new measures of influence, based on prior research, extended appropriately for recommender systems, and (d) the feasibility and implications of meaningful applications of influence.