Polarization isn’t just a matter of individual beliefs — it also emerges from how we interact and form relationships. By looking at how social networks form and how groups talk to each other at scale, these studies show that division and hostility follow predictable patterns.
Online Discourse is Falsely Polarized by Self-Selection
William Small Schulz, Claire E. Robertson, Hallgeir Sjåstad, Ashwini Ashokkumar, Lisa Oswald, JungHwan Yang, Simon Munzert, Jay J. Van Bavel & Andrew M. Guess. Preprint, September 2026.
Are we really as polarized as we seem? Across 1.66 million posts and a survey of 40 countries, people with strong views posted more than moderates — and people posted their own extreme opinions more than their moderate ones. This makes polarization seem worse online than in reality.
Emergence of polarization in networks of large language model agents
Jinghua Piao, Zhihong Lu, Chen Gao, Fengli Xu, Qinghua Hu, Fernando P. Santos, Yong Li & James Evans. Nature Communications, October 2026.
When thousands of AI agents interacted in simulated social networks, they sorted themselves into like-minded clusters and became polarized, much like people do.
Unifying Models of Intergroup Hostility in Online Discourse
Patrick Gerard, Julia Mendelsohn & Kristina Lerman. Preprint, September 2026.
An analysis of over 2 million posts from TikTok, Truth Social and Twitter/X during the 2024 election found that hostile rhetoric toward social groups tends to rely on “us vs. them” language and on framing the other group as a threat, forming a predictable pattern.