Opinion Homogenization and Polarization

We describe three sampling models that aim to cast light on how some design features of social media platforms systematically affect judgments of their users. We specify the micro-mechanisms of belief formation and interactions and explore their macro implications such as opinion polarization. Each model focuses on a specific aspect of platform-mediated social interactions: how popularity creates additional exposure to contrarian arguments; how differences in popularity make an agent more likely to hear particularly persuasive arguments in support of popular options; and how opinions in favor of popular options are reinforced through social feedback. We show that these mechanisms lead to self-reinforcing dynamics that can result in local opinion homogenization and between-group polarization. Unlike nonsampling-based approaches, our focus does not lie in peculiarities of information processing such as motivated cognition but instead emphasizes how structural features of the learning environment contribute to opinion homogenization and polarization.