The Logical Dynamics of Social Networks: From Homophily to Polarization
Jonathan Thul

Abstract:
This thesis studies the effect of homophily on the development of social networks. Homophily refers to the tendency of individuals to be socially connected to others that are similar to themselves. We contend that this phenomenon can be traced back to two tendencies, social influence and social selection. Using techniques from dynamic epistemic logic, we provide a formal setting to represent social networks and define model transforming upgrades that correspond to social influence and social selection.
 In the first part of the thesis, we introduce the notion of cluster-split models and argue that they represent socially fragmented networks. We show how social selection gives rise to such models and argue that this suggests a connection between homophily and polarization.
 In the second part we introduce epistemic social network models. This allows us to define different epistemic update versions of social influence and social selection. We compare these updates, showing that the different ways agents deal with epistemic uncertainty can lead to different network developments. Finally, we survey phenomena of learning that arise in our framework.