Three Essays on Trust Mining in Online Social Networks

Abstract

This dissertation research consists of three essays on studying trust in online social networks. Trust plays a critical role in online social relationships, because of the high levels of risk and uncertainty involved. Guided by relevant social science and computational graph theories, I develop conceptual and predictive models to gain insights into trusting behaviors in online social relationships. In the first essay, I propose a conceptual model of trust formation in online social networks. This is the first study that integrates the existing graph-based view of trust formation in social networks with socio-psychological theories of trust to provide a richer understanding of trusting behaviors in online social networks. I introduce new behavioral antecedents of trusting behaviors and redefine and integrate existing graph-based concepts to develop the proposed conceptual model. The empirical findings indicate that both socio-psychological and graph-based trust-related factors should be considered in studying trust formation in online social networks. In the second essay, I propose a theory-based predictive model to predict trust and distrust links in online social networks. Previous trust prediction models used limited network structural data to predict future trust/distrust relationships, ignoring the underlying behavioral trust-inducing factors. I identify a comprehensive set of behavioral and structural predictors of trust/distrust links based on related theories, and then build multiple supervised classification models to predict trust/distrust links in online social networks. The empirical results confirm the superior fit and predictive performance of the proposed model over the baselines. In the third essay, I propose a lexicon-based text mining model to mine trust related user-generated content (UGC). This is the first theory-based text mining model to examine important factors in online trusting decisions from UGC. I build domain-specific trustworthiness lexicons for online social networks based on related behavioral foundations and text mining techniques. Next, I propose a lexicon-based text mining model that automatically extracts and classifies trustworthiness characteristics from trust reviews. The empirical evaluations show the superior performance of the proposed text mining system over the baselines.