When Words Divide: Diachronic Ideological Polarization in Political Discourse on Social Media

2026-08-02

AI Summary

This study addresses the lack of systematic understanding regarding the mechanisms driving the evolution of ideological polarization in long-term online political discourse. To overcome the limitations of cross-sectional or sentiment-focused analyses, this work proposes a temporally aligned, community-specific word embedding approach that integrates semantic distance metrics with large-scale Reddit discussion data. For the first time, this framework enables longitudinal modeling of semantic divergence between opposing political groups on key concepts. Empirical results demonstrate a significant intensification of ideological polarization at both conceptual and topical levels over the study period, confirming the method’s capacity to effectively capture dynamic polarization processes in large-scale textual corpora.

Abstract

Political polarization has become a defining feature of online discourse, yet its long-term evolution remains poorly understood. We present a longitudinal analysis of ideological polarization in Reddit discussions by measuring semantic differences in the language used by opposing political communities. We construct temporally aligned community-specific word embeddings and quantify ideological polarization as the semantic divergence of political concepts over time. Our analysis shows that ideological polarization has increased substantially during the study period, both at the concept- and topic-level. Unlike prior computational work, which has largely focused on cross-sectional analyses or affective dimensions of polarization at a single point at time, our approach captures the evolution of ideological differences in semantic framing. The proposed framework provides a scalable method for studying the temporal dynamics of ideological polarization in large-scale social media discourse.

Problem

• ideological polarization
• diachronic analysis
• semantic divergence
• political discourse
• social media

Innovation

• temporally aligned word embeddings
• semantic divergence
• ideological polarization
• longitudinal analysis
• political discourse