Computational Trust at Various Granularities in Social Networks

Trust has been a ubiquitous phenomenon in human lives. The phenomenon of trust has been studied at various granularities over the centuries by various researchers encompassing all disciplines of academia. Historically, it has been witnessed that the primary mode of studying trust has been surveying subjects and documenting the results. But the burgeoning electronic social media have provided us with the unique opportunity of studying trust under a new perspective, which is known as computational trust. Computational trust is defined as the generation of trust between two human actors mediated through computers. This is an active area of research due to the proliferation of various socially rich datasets over the past decade. This includes massively multi-player online games (MMOs), online social networks and various web services, allowing actors to trust each other in an online virtual setting.;The first part of this thesis investigates various aspects affecting dyadic (or interpersonal) trust, i.e., trust between two actors. This includes formation, reciprocation and revocation of trust. Taking into account various nuances of dyadic trust, this thesis predicts the occurrence of these three phenomena in the datasets. Instead of looking at these phenomena by itself, this thesis looks at this phenomena in conjunction with social relations for better predictive modeling. One of the major requirements in trust applications is identifying the trustworthy actors in the social networks which will be the subject of investigation for the second part of this dissertation. An important factor in the prediction of trust is an actor's inherent ability to trust others and the perception of the actor in the network. This thesis proposes a pair of complementary measures that can be used to measure trust scores of actors in a social network using involvement of social networks. Based on the proposed measures, an iterative matrix convergence algorithm is developed that calculates the trustingness and the trustworthiness of each actor in the network. Trustingness of an actor is defined as the propensity of an actor to trust his neighbors in the network. Trustworthiness, on the other hand, is defined as the willingness of the network to trust an individual actor. The algorithm runs in O(k x |E|) time where k denotes the number of iterations and |E| denotes the number of edges in the network. This thesis also shows that the algorithm converges to a finite value very quickly. Lastly, this thesis introduces the concept of "vulnerable paths" and identifies those paths in a social network. Based on the hypothesis that these vulnerable paths are imperative for influence flow, a new algorithm proposed in this thesis, exploits these paths for better and more targeted viral marketing using trust scores. It is shown that there is an improvement as high as 9% in identifying these paths using the proposed algorithm than state of the art trust scoring algorithms.;This thesis makes the following contributions. It studies the generative mechanisms of trust not in isolation, but in conjunction with the social processes(relations) around trust. Whereas earlier studies were interested in looking at the cross-sectional view of trust, this study investigates the longitudinal view of trust. Instead of looking only at the dynamics of initiation of interpersonal trust, this study looks at the various other dynamics such as reciprocation and revocation of interpersonal trust. This study also exploits the negative feedback property in trust to propose computationally stable pair of global trust measures, which can be used to measure the propensity of actors to trust and be trusted in a network. Finally, this pair of scores is leveraged to be used in various applications such as viral marketing, identification of "vulnerable paths" and inoculation of a network from rumor spread.