Designing Mechanisms for Trust-based Interaction in Social Networks

In this thesis, we build an agent-based framework to model trust-based interaction among people in a social network. We assume that trust is one person’s expectancy that another person will behave in a way that a good outcome for both people is achieved. The social network is represented by a weighted graph in which the nodes are the agents and the links reflect the strength of trust relationships between agents. Agents interact with each other – for example, by exchanging recommendations or forming coalitions. Based on feedback mechanisms that take the utility which agents perceive into account, agents dynamically update their trust relationships to other agents. Our motivation to study trust-based interaction in social networks comes from the facts that, on the Internet, people are increasingly confronted with information overload as well as information asymmetries. By focusing on the social network and the trust relationships between people in communities on the Internet, it is possible to (a) design mechanisms of trust-based interaction on social networks that achieve a trade-off between individual utility and social efficiency and to (b) better personalise a vast range of services provided on the Internet. First, we present a model of a trust-based recommender system on a social network whose main idea is that users can use their social network to reach information and their trust relationships to filter the information to separate the relevant and the irrelevant. We investigate how the dynamics of trust among agents affect the performance of the system and compare it to a frequency-based recommender system. We identify the impact of network density, preference heterogeneity among agents, and knowledge sparseness on the performance of the system. The system self-organises to a state with performance near to the optimum; the performance on the global level is an emergent property of the system, achieved without explicit coordination from the local interactions of agents. Second, we propose a novel trust metric for social networks – we refer to it as the TrustWebRank metric. It is personalised and dynamic, and allows us to compute the indirect trust between two users which are not neighbours based on the direct trust between users that are neighbours. By analogy to PageRank, this metric makes use of the concept of feedback centrality and overcomes some of the limitations of other trust