Controversy-Aware Hybrid Trust Inference in Online Social Networks

Social interactions assume a central role in everyday life of people strongly connected through online social networks and web communities. Trust can be used as a filter to the increasing flow of information, experiences and opinions exchanged among users. In the absence of direct relations, trust inference algorithms can be exploited in order to estimate missing trust values. However, controversial situations can easily arise from the subjectivity of trust opinions, when the lack of agreement leads to ambiguity and uncertainty in the inferred trust. In this work, we propose a novel trust inference algorithm, named TrustZic, which aims at improving the performance of trust inference in controversial conditions by combining local and global trust aspects. The proposed approach is based on the use of global reputation values as node weights biasing a local trust inference process. Experimental evaluation has shown the significance of our approach and its effectiveness in supporting trust decisions in critical situations characterized by a high level of controversy.