Emergence of polarization in networks of large language model agents

Abstract Rapid advances in large language models (LLMs) have empowered autonomous agents to generate social networks, communicate, and form shared and diverging opinions on political issues. However, our understanding of their collective behaviours and underlying mechanisms remains incomplete. In this paper, we simulate networked systems involving thousands of LLM agents across different backbone models (GPT-3.5, GPT-4o, ChatGLM, Llama-3, and DeepSeek-V3), in which agents interact through LLM-guided conversations and update their opinions over time, resulting in the emergence of opinion polarization. We discover that these agents spontaneously develop their own social network with properties characteristic of human social networks, such as homophilic clustering. The collective opinions of these LLM agents evolve in ways that exhibit behavioural patterns consistent with social phenomena and mechanisms widely discussed in empirical studies of human behaviour and opinion-dynamics models. This consistency suggests that LLM agents can serve as a valuable synthetic testbed for exploring hypothetical intervention strategies in networked LLM-agent systems. Using this testbed, we further examine the effects of a range of network- and individual-level interventions, such as promoting more diverse interactions and reducing confirmation bias. Overall, this work not only sheds light on subtle opinion dynamics and collective behaviour of LLM agents from a network perspective, but also demonstrates their potential for testing hypotheses about opinion dynamics and intervention strategies.