Social groups characterization and dynamics: a network science approach

Abstract

Over the course of our lives, we tend to transition through many social groups. From an early age, our first social group is our family, then we have our schoolmates, later in life, our coworkers, and likely another family. As we move from one group to another, we embed ourselves in the dynamics of social groups’ losses and gains of new members. These dynamics are a complex system of interactions, which at scale, form the structural basis of our societies. In many years of sociological research, the details of social group interactions remained poorly understood. Certainly, not because of lack of importance, as they play a central role in society, but rather because of the inherent complexity of the phenomena as well as the scarcity of tools and data necessary to tackle the problem. The coming-of-age of computing technology revolutionized the way we interact and communicate. As our societies become more digitalized, we generate pieces of information which are traces of our daily lives and social interactions. These traces contain the essence of our social structures and represent the means of approaching previously unaddressable problems. In this dissertation, we leverage the power of data and computation to uncover the patterns of group to group interaction and shed light on the dynamics of social groups’ affiliation exchange. Our contribution is as follows, (1 ) first, we designed an information-centered approach to evaluate the existing methods of identification of social groups in network data; (2 ) then, we show that we can use affiliation exchange in the context of politics to infer the ideological leaning of certain political groups, by modeling changes in ideology as a function of affiliation information; (3 ) and finally, we characterize the properties of group-group interactions of academic affiliations and the implications in that specific context.