Analysis of the Structure of Social Networks for Information Diffusion

The vast proliferation of Online Social Networks (OSN) is generating many new ways to interact and create social relationships with others. In OSN, information spreads among users following existing social relationships. This spread is influenced by the local properties and structures of the social relationships at individual level. Being able to understand these properties can be fundamental for the design of new communication systems able to predict the creation and sharing of content based on social properties of the users. While substantial results have been obtained in anthropology literature describing the properties of human social networks, a clear understanding of the properties of social networks built using OSN is still to be achieved. In this thesis, the structure of Ego networks formed online is compared with the properties of offline social relationships showing interesting similarities. These properties are exploited to provide a meaningful way to study the mechanisms controlling the formation of information diffusion chains in social networks (typically referred to as information cascades). Trough the analysis of synthetically generated diffusion cascades executed in a large Facebook communication datasets, is showed that the knowledge of tie strength of the social links is fundamental to infer which nodes will give rise to large information cascades and which links will be more used in the information diffusion process. We analysed the trade off between information spread and trustworthiness of information. Specifically, we have investigated the spread of information when only links of a certain trust value are used. Assuming, based on results from sociology, that trust can be quantised, we show that too strict limits on the minimum trust between users limit significantly information spread. In the thesis we investigate the effect of different strategies to significantly increase spread of information by minimally relaxing constraints on the minimum allowed trust level.