A wider picture of trust: confronting multi-contextual social traits for trust inference

Abstract

With the advent of social networks, problems related to information veracity and people trustworthiness naturally arise and become more acute as the numbers of participants and the volumes of exchanged information grow. Predicting and recommending trust (and distrust) levels is essential for a healthier online presence. Although many research efforts are being devoted to these tasks, most studies that have been undertaken in this area are constrained by two facts. 1) Distrust, contrary to trust, is not necessarily propagative, thus is harder to infer using propagative approaches. 2) Most publicly available datasets fail to capture the essence of trust for they ignore one of its most important aspects: its context. In this paper, we propose a multidimensional trust model that allows us to describe some social traits that affect trust, and which could not be described with a unidimensional model. We, then, use these traits to propose a straightforward trust prediction approach that does not rely on trust propagation. Our work aims to provide a usable foundation and to ignite more advanced researches on trust and its applications.