On Selecting Recommenders for Trust Evaluation in Online Social Networks

Trust is a central component of social interactions among humans. Many applications motivate the consideration of trust evaluation in online social networks (OSNs). Some work has been proposed based on a trusted graph. However, it is still an open challenge to construct a trusted graph, especially in terms of selecting proper recommenders, which can be used to predict the trustworthiness of an unknown target efficiently and effectively. Based on the intuition that people who are close to and influential to us can make more proper and acceptable recommendations, we present the idea of recommendation-aware trust evaluation (RATE). We further model the recommender selection problem as an optimization problem, with the objectives of higher accuracy, lower risk (uncertainty), and lower cost. Four metrics: trustworthiness, expertise, uncertainty, and cost, are identified to measure and adjust the quality of recommenders. We focus on a 1-hop recommender selection, for which we propose the FluidTrust model to better illustrate the trust--decision making process of a user. We also discuss the extension of multihop scenarios and multitarget scenarios. Experimental results, with the real social network datasets of Epinions and Advogato, validate the effectiveness of RATE: it can predict trust with higher accuracy (it gains about 20% higher accuracy in Epinions), lower risk, and less cost (about a 30% improvement).