Harnessing Facebook for the evaluation of recommender systems based on physical copresence

Abstract

Various mobile social applications have proposed the use of ad-hoc network connectivity as a means of detecting user encounters and shared social contexts. These applications range from simple opportunistic information sharing to techniques for collaborative filtering in mobile settings. However, it can be difficult and costly to test the underlying assumption that repeated physical copresence can be used as a measure of user similarity. We have therefore developed a framework that allows existing online social platforms such as Facebook to be coupled with simple, standard mobile applications in order to test such hypotheses. The central idea is to map the physical copresence of users to connections in virtual social networks and then exploit the rich support for developing pluggable applications to measure user similarity within these networks.