Social networks: modeling structure and dynamics

Abstract

The study of networks of social interaction can be seen to originate from the work of Jacob Moreno in the 1920's. At the turn of the millennium new actors entered the field, researchers with a background in physics and computer science, who brought with them a new set of tools that could be used to collect and analyse large sets of data. Analysis of large scale social network data from various sources has increased our knowledge of the common features of various social networks, observed in networks of acquaintance and collaboration alike. The quantification and modeling of a particular feature of social networks, namely the tendency of individuals to form densely connected groups with relatively few links to individuals outside the group (called communities in complex networks theory), has taken large steps in recent years. Modeling these structures and their effect on social dynamics is a highly topical issue, relevant for fields such as spreading of epidemics or rumors and formation of opinions, with applications such as prevention of epidemics and marketing. This thesis aims to increase our understanding of the structure of large scale social networks, and of dynamics unfolding in such networks, in several ways: 1) In order to answer a need for social network models that generate realistic structures at large scale, we introduce a model based on simple local mechanisms leading to community structure. 2) A thorough comparative study of models for social networks assesses the adaptability of the models to fit real social network data, and their success at reproducing prominent structural features of social networks. In discussing in detail two major approaches to modeling social networks, this study may promote the understanding between researchers from the two 'schools of thought'. 3) We study models of competing options, with focus on perhaps the most important feature of social network structure, namely communities, that had been largely lacking in earlier research.