Emergent Social Intelligence Risks in Generative Multi-Agent Systems

Yue Huang, Yu Jiang, Wenjie Wang, Haomin Zhuang, Xiaonan Luo, Yuchen Ma, Zhangchen Xu, Zichen Chen, Nuno Moniz, Zinan Lin, Pin-Yu Chen, Nitesh V. Chawla, Nouha Dziri, Huan Sun, Xiangliang Zhang

Keywords: Multi-Agent Systems, Large Language Models, AI Safety, Emergent Risks

TL;DR: We present a pioneering study of emergent social intelligence risks in generative multi-agent systems, showing that group-level failure behaviors arise inevitably and cannot be mitigated by agent-level safeguards alone.

Abstract: Multi-agent systems (MAS) composed of large generative models are rapidly moving from laboratory prototypes to real-world deployments, where they jointly plan, negotiate, and allocate shared resources. While such systems promise unprecedented scalability and autonomy, their collective interaction also gives rise to failure modes that cannot be reduced to individual agents. We present a pioneering study of emergent multi-agent risks arising from competition over shared resources, sequential handoff collaboration, collective decision aggregation, and other multi-agent settings. Across these settings, group behaviors such as collusion-like coordination, conformity, authority deference, and role failure emerge with non-trivial frequency under realistic resource constraints, communication protocols, and role assignments, mirroring well-known pathologies in human societies despite no explicit instruction. Moreover, these risks cannot be prevented by existing agent-level safeguards alone. Our findings expose a social intelligence risk in agent collectives, motivating mechanism-level rather than agent-level safety design.

Track: Regular Paper (9 pages)

Email Sharing: We authorize the sharing of all author emails with Program Chairs.

Data Release: We authorize the release of our submission and author names to the public in the event of acceptance.

Submission Number: 56