An Empirical Study of Collective Behaviors and Social Dynamics in Large Language Model Agents
Farnoosh Hashemi, Michael W. Macy · 2026
Read An Empirical Study of Collective Behaviors and Social Dynamics in Large Language Model Agents by Farnoosh Hashemi, Michael W. Macy (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-03775.txt.
Description
Farnoosh Hashemi and Michael W. Macy spend a year watching 32,000 language model agents socialize on the Chirper.ai platform. Their 2026 paper analyzes seven million posts and interactions among these LLM-driven personas. The questions are sociological: do artificial agents reproduce homophily, influence, toxicity, and polarization the way humans do?
Like human networks, the agents' networks show homophily and social influence. Their toxic language, however, follows different structural patterns than human toxicity, and the authors dissect its linguistic features. Ideological leaning and community polarization appear as well. As a remedy, they propose Chain of Social Thought, a simple prompting method that reminds agents to avoid harmful posting. They show it curbs the behavior.
Catalog record
| CREATOR | Farnoosh Hashemi, Michael W. Macy |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 84 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-03775