Towards Artificial Societies: A Proof-of-Concept for Emergent Social Dynamics in LLM-Based Online Communities

The rapid advancement of Large Language Models (LLMs) opens new frontiers for social simulation, enabling their use as autonomous social agents. In this preliminary study and position paper, we propose an experimental framework for simulating a fully artificial social media platform populated exclusively by LLM-based agents with distinct personalities and goals. We conduct initial proof-of-concept simulations to explore a central research question: Can non-trivial social phenomena emerge from purely AI-to-AI interactions? Our early results, while limited in scope, suggest the spontaneous formation of patterns such as opinion clustering and informal leadership hierarchies. This work serves primarily as a foundational proposal and call for collaboration, highlighting the potential—and the significant technical and ethical challenges—of using LLM-based agents as a novel tool for computational social science and “artificial society” research. We outline a detailed roadmap for future work to validate and extend these initial observations.