Detecting social identities through network analysis and agent-based modelling

Reciprocal, social interactions create links between people, such as the expression of opinions, beliefs and attitudes, including on online platforms like Twitter. Attitude networks where people are linked by their attitude similarities reveal the foundations of potential groups that reflect social identities. Homophily drives the social structure in these attitude networks, in which groups form around common opinions.

Although opinion-based groups define individuals’ identities and shape their attitudes, the detection and analysis of opinion-based groups have only been investigated in a limited manner. The objective of this thesis is to enhance our understanding of social structures. Therefore, we investigate whether opinion-based groups can be identified as meaningful structures within attitude networks. Exploring the structural relations between these groups discloses social fragmentation, such as polarisation.

Large, representative survey data on attitudes, e.g., towards political identities, provide a snapshot of society’s status quo. We transform survey data into attitude networks to reveal structural properties of groups and social identities. While we utilise network analysis for identifying group formation and shifts in attitude alignments, agent-based modelling provides us with a way to mimic the dynamics of attitude formation that are driven by the reciprocal influence of individuals. The analysis of networks from survey data opens up a multitude of innovative possibilities for further investigation. Structural properties of the attitude networks allow evaluations on within-group cohesion and inter-group relations. Opinion-based group detection is linked through attitude alignment to the evaluation of ideological polarisation. From attitudinal survey data from the United States, we uncover and evaluate group structures as well as trends in polarisation. By modelling the development of attitudes through agent-based models, we provide the requirements and explore the limitations for predicting the evolution of attitude networks.

Through defining people’s behaviours, emotions and attitudes, social identities frame societal dynamics. With our analysis of attitude networks as well as the exploration of attitude changes through agent-based modelling, we identify patterns of emerging social identities