Quantifying Multipolar Polarization

TL;DR

This paper compares methods for measuring multipolar polarization in social networks, finding that an extended Euclidean distance metric effectively captures complex multi-opinion systems, improving over traditional two-dimensional models.

Contribution

It introduces and empirically evaluates an extension of Euclidean distance for quantifying multipolar polarization, addressing limitations of existing two-dimensional approaches.

Findings

The average pairwise distance extension performs well on desired properties.

It effectively models multi-opinion polarization systems.

Provides a practical metric for complex social network analysis.

Abstract

Studying and understanding social networks is crucial for accurately defining ideological polarization, since they enable precise modeling of social structures. One of the limitations of many methods for quantifying polarization on networks is the assumption of a two-dimensional opinion space. This prevents accurate study of multipolar systems like multi-party political systems, where modeling more than two opinion poles is beneficial. Here, I experimentally compare methods for quantifying multipolar polarization on a network and find that the average pairwise distance extension of generalized Euclidean distance conforms to several desired properties, showing its advantages over other methods. This allows the study of multipolar polarized systems based on an empirically and intuitively good metric.