Polarisation and Homophily: Experiments with a Contagion-Repulsion Combination Opinion Dynamics Model

To address social concerns that online personalisation algorithms increase polarisation, scholars have studied two distinct classes of agent based models of opinion dynamics constructed to explain polarisation: contagion/reinforcement class and repulsion class. It has been rigorously demonstrated that these classes make contrasting predictions on the relationship between polarisation and homophily (thought to be a consequence of online personalisation algorithms). However, consequences of a model that combines contagion and repulsion on the relationship between homophily and polarisation has not been rigorously studied. Thus, this paper proposes a theoretical foundation for such a model and extends an existing contagion class attitude dynamics model into a contagion-repulsion combination model. This results in a novel equation of attitude change which positions the combination model as a generalisation of several other opinion dynamics models. Subsequently, simulation experiments test implications of this model on the relationship between homophily and polarisation. Results show different parameter combinations produce several qualitatively different forms of the relationship, ranging from increasing, decreasing, U-shaped or even flat. This suggests the relationship between polarisation and homophily is sensitive to context and may be issue dependent.