The Polarization Paradox: Why More Connections Can Divide Us

Abstract

We develop a simple model to examine how social network structure shapes political polarization. Despite its tractability, the model generates a rich set of equilibrium behaviors. First, we show that moderate views, expressed by centrist contents, benefit from a pairwise advantage and is amplified when it is the most common type. However, when extreme views dominate, sufficiently dense networks can amplify polarization. Second, we establish a non-monotonic relationship between network density and polarization, offering a theoretical explanation for the mixed empirical evidence on the role of social media. Third, we show that partial polarization is less likely in nearly regular networks than in those with more dispersed degree distributions. We also explore two extensions: one where agents can abstain from forwarding extreme content, and another where ideological preferences are asymmetric. In both cases, the results remain, but the conditions for full or no polarization become more restrictive.