How Polarization Extends to New Topics: An Agent-Based Model Derived from Experimental Data

Polarization is a key phenomenon which has been linked to increasing disliking between people of opposite political groups. Furthermore, polarization can extend to new topics such as the debate on COVID-19 vaccines, making it more complex to coordinate efforts for such a problem. The social identity approach (SIA) offers a robust theoretical framework for understanding identity-based social processes. This approach suggests that people's perceptions and behaviour depend on their group identity (e.g., Democrat vs Republican). In this ar-ticle, we developed an opinion-dynamics model integrating SIA to explore how polarization can extend to new topics. Furthermore, we developed this model from experiments with human participants. This allows us to use already validated micro-dynamic rules in the model. Empirical results show lack of repulsive effects, more attraction during in-group interactions and a new effect: increased stubbornness when people are exposed to opinions of an out-group member. The model was built mimicking the interaction structure of the experiment. At each iteration, an agent observes the opinion of another agent. Depending on their respective groups the agent will experience a stronger or weaker attractive force, together with some noise. This model was able to produce polarization without the use of repulsive forces. Furthermore, the sensitivity analysis tells us that po-larization in new topics can appear when all the following conditions are satisfied: (1) each person recognizes who is belonging to which political group, (2) there are more in-group than out-group interactions and (3) there is some initial asymmetry on the topic.