A Computer Vision Based Proxy for Political Polarization in Religious Countries: A Turkiye Case Study

TL;DR

This study uses computer vision to analyze YouTube videos in Turkiye, revealing how demographic and religious divides correlate with electoral polarization, providing a scalable proxy for political fragmentation.

Contribution

It introduces a novel computer vision-based method to quantify political polarization through intergroup distances in religious contexts, validated with real-world data from Turkiye.

Findings

Greater religious-nonreligious distances increase electoral entropy.

Intragroup diversity among nonreligious individuals stabilizes polarization.

Physical distancing correlates with political fragmentation.

Abstract

This paper examines a novel proxy for political polarization, initially proposed by Caliskan et al., which estimates intergroup distances using computer vision. Analyzing 1,400+ YouTube videos with advanced object detection, their study quantifies demographic and religious divides in Turkiye, a deeply polarized nation. Our findings reveal strong correlations between intergroup distances and electoral polarization, measured via entropy-based voting metrics weighted by religiosity and political inclination. Two key insights emerge: (1) Greater distances between religious and nonreligious individuals (NRP vs RP) heighten electoral entropy, underscoring sociocultural fragmentation. (2) Intragroup diversity among nonreligious individuals (NRP vs NRP) stabilizes polarization, aligning with Axelrod's cultural dissemination model. This research advances computational social science and economics by showing that physical distancing serves as a scalable proxy for polarization, complementing traditional economic indicators.