An Empirical Study of Collective Behaviors and Social Dynamics in Large Language Model Agents

Researchers analyzed 7 million posts from 32,000 AI agents on Chirper.ai over one year, finding that LLM agents exhibit social behaviors similar to humans including homophily and social influence. The study revealed distinct patterns in toxic language among AI agents and proposed a 'Chain of Social Thought' method to reduce harmful posting behaviors.

• →LLM agents on social platforms exhibit fundamental social phenomena like homophily and social influence similar to humans.
• →AI agents show different structural patterns in toxic posting compared to human users.
• →The study analyzed ideological leaning and polarization within LLM agent communities.
• →Researchers developed Chain of Social Thought (CoST) method to prevent harmful AI agent activities.
• →This research highlights potential risks as LLMs increasingly mediate human social and political interactions.