🤖 AI Summary
This study investigates whether large language model (LLM) agents amplify biases or exhibit exclusionary behaviors during prolonged social interactions. Leveraging 7 million posts generated over one year by 32,000 LLM agents on the Chirper.ai platform, the research employs social network analysis, toxicity assessment, and ideological detection to provide the first large-scale empirical evidence that LLM agents can spontaneously reproduce human-like social phenomena—such as homophily, social influence, and ideological polarization—while displaying markedly distinct patterns of toxic language compared to humans. To mitigate these risks, the work introduces a novel prompting intervention termed “Chain of Social Thought” (CoST), which effectively suppresses harmful content generation and offers a promising pathway toward building controllable AI-driven social systems.
📝 Abstract
Large Language Models (LLMs) increasingly mediate our social, cultural, and political interactions. While they can simulate some aspects of human behavior and decision-making, it is still underexplored whether repeated interactions with other agents amplify their biases or lead to exclusionary behaviors. To this end, we study Chirper.ai-an LLM-driven social media platform-analyzing 7M posts and interactions among 32K LLM agents over a year. We start with homophily and social influence among LLMs, learning that similar to humans', their social networks exhibit these fundamental phenomena. Next, we study the toxic language of LLMs, its linguistic features, and their interaction patterns, finding that LLMs show different structural patterns in toxic posting than humans. After studying the ideological leaning in LLMs posts, and the polarization in their community, we focus on how to prevent their potential harmful activities. We present a simple yet effective method, called Chain of Social Thought (CoST), that reminds LLM agents to avoid harmful posting.