LLMs Trust Their Own: Identity-Dependent Conformity in Multi-Agent Systems

2026-09-27

AI Summary

This study investigates whether large language models (LLMs) in multi-agent systems exhibit conformity bias driven by social identity rather than answer correctness. To this end, we construct a multi-agent judgment task environment incorporating differentiated social identities and systematically evaluate the conforming behaviors of twelve open-source LLMs under various reasoning strategies. Our findings reveal an identity-driven bidirectional effect: in-group consensus significantly amplifies conformity, whereas out-group consensus suppresses it. Notably, chain-of-thought reasoning fails to fully mitigate in-group influence. This work demonstrates that social identity shapes information aggregation mechanisms independently of factual correctness, thereby identifying a critical security attack surface in multi-agent systems and providing essential empirical evidence for AI safety alignment.

Abstract

Large language models (LLMs) are increasingly deployed in multi-agent settings, where agents observe and influence one another, making social influence a key dimension of AI behavior and safety. We investigate whether LLMs'responses depend on the social identity of other agents, beyond the effect of their consensus. We construct judgment tasks with a single correct answer, and place models in a multi-agent setting where they receive incorrect answers from other agents whose social identities (AI or human, model family, or an arbitrary minimal group) are either shared with or distinct from their own. Across 12 open-weights models and nine tasks, we find a bidirectional effect of group identity on conformity to incorrect answers: in-group consensus increases conformity (in-group favoritism), whereas out-group consensus decreases it (out-group divergence). Unlike humans, for whom one ally breaking the consensus sharply reduces conformity, models are unmoved by an ally from the majority's group. Worse, a correct ally from the opposing group intensifies this bidirectional effect. Chain-of-Thought reasoning suppresses most of these effects, yet an in-group ally still reduces conformity to an incorrect out-group majority. Labeling peers as safety-aligned shifts overall conformity but leaves in-group favoritism and out-group divergence intact. These results show that group identity shapes how LLMs aggregate information across agents, independently of its correctness, and identify a manipulation surface for multi-agent AI systems.

Problem

• Large Language Models
• Multi-Agent Systems
• Social Identity
• Conformity
• In-group Favoritism

Innovation

• Multi-Agent Systems
• Social Identity
• Conformity
• Large Language Models
• Chain-of-Thought