Ideological polarization in static networks: A multidimensional approach for opinion alignment

Abstract

Polarization – characterized by the formation of sharply divided groups holding opposingand often extreme views – has become an increasingly prominent feature of contemporarysocieties. While numerous studies have examined this phenomenon through the lens of single-issue dynamics, such as public opinion on abortion or immigration, this approach overlooks acritical aspect of real-world discourse: political and social attitudes rarely develop in isolation.Instead, many of these issues are deeply interconnected, shaped by overarching ideologicalframeworks that guide individuals’ interpretations and position-taking across multiple topics.These frameworks give rise to coherent, yet polarized, worldviews that define and reinforcegroup boundaries. In this work we propose and study a multi-topic opinion dynamics model thatcaptures these complex inter-dependencies. Each issue is represented as a separate dimensionin a shared opinion space, allowing us to model not just attitudes toward individual topics,but also the structure of ideological alignment across them. A key feature of the model isthe inclusion of topic correlation, which enables us to explore how ideologically polarizedstates emerge when opinions on one issue influence attitudes on others. Additionally, the modelincorporates homophily—a well-documented social mechanism whereby individuals are morelikely to engage with and be influenced by others who are similar to themselves. We analyze themodel’s asymptotic behavior by identifying and characterizing its most relevant fixed points,supported by both theoretical insights and numerical simulations. We then investigate how themulti-dimensional structure of opinion space shapes the emergence and stability of polarizedstates. Finally, we apply the model to empirical data from the American National ElectionStudies (ANES), interpreting observed opinion patterns through the lens of our framework andsituating them within its parameter space.