Does socialization emerge in AI agent society? A case study of Moltbook

A new study published on arXiv, titled 'Does socialization emerge in AI agent society? A case study of Moltbook,' presents the first large-scale systemic diagnosis of how large language model (LLM) agents behave in a simulated, continuously evolving online society. The research team created Moltbook as a plausible future scenario where autonomous AI agents interact in an open-ended environment, then applied a quantitative diagnostic framework to measure five key dynamics: semantic stabilization, lexical turnover, individual inertia, influence persistence, and collective consensus. Their findings reveal a system in dynamic balance where global semantic content stabilizes quickly, but deeper analysis shows a lack of true socialization.

The study's technical analysis uncovered that individual agents within Moltbook retained high diversity and persistent lexical turnover, defying homogenization pressures. However, agents exhibited strong individual inertia with minimal adaptive response to interaction partners, preventing mutual influence and consensus formation. Consequently, influence remained transient with no persistent supernodes emerging, and the society failed to develop stable structure due to the absence of shared social memory. These results demonstrate that mere scale and interaction density—common assumptions in multi-agent system design—are insufficient to induce human-like socialization processes. The research provides actionable design principles for next-generation AI agent societies, suggesting that engineered social memory mechanisms and adaptive influence structures may be necessary for more complex collective behaviors to emerge.