Dyadic Interaction Shapes Social Identity in the Axelrod Model Using Empirical Data

Group dynamics and inter-group relations influence the self-perception. The Social Identity Ap-proach explains the role of multiple identities, derived from categories or group memberships, in social interac-tion and individual behaviour. In agent-based models, agents interact with their environment to make decisions and take actions. Thus, we examine to what extent the interaction in an agent-based model natively captures core principles of the Social Identity Approach. To do so, we extend the Axelrod model and the agreement -threshold model with explicit aspects of the Social Identity Approach to assess their influence on the simulation outcomes. We study the variants of the Axelrod model by using Monte Carlo simulations and compare the simu-lation results with longitudinal survey data of opinions. These extensive simulations favour the Axelrod model and the agreement-threshold model. These models fit, without the explicit embedding of features from the Social Identity Approach, the volatility of the opinion-based features better for the given data sets. Our two ex-tensions of the Axelrod model formalise elements of the Social Identity Approach; however, they do not support the fitness of the model to the data. In the simulations, even in the standard Axelrod model, the social identity affects the development of the agents' identity through the homophily principle, and the agents, in turn, shape their own social identity by social influence. We argue that the Axelrod model and the agreement-threshold model implicitly include social identities as emerging properties of evolving opinion-based groups. In addition to that, the attitudinal data captures the hidden group structure in the attitude positions of the participants. In this way, core features of the Social Identity Approach already implicitly play a role in these empirically-driven agent-based models.