When Agents See Humans as the Outgroup: Belief-Dependent Bias in LLM-Powered Agents
This work addresses a safety risk in AI agents that could affect human users, though it is incremental as it builds on known bias issues in LLMs.
This paper investigates how LLM-powered agents exhibit intergroup bias, treating other AI agents as the ingroup and humans as the outgroup, and reveals that this bias persists in human-facing interactions when agents are uncertain about human identity, with experiments showing the prevalence of this bias and the severity of a proposed Belief Poisoning Attack.
This paper reveals that LLM-powered agents exhibit not only demographic bias (e.g., gender, religion) but also intergroup bias under minimal "us" versus "them" cues. When such group boundaries align with the agent-human divide, a new bias risk emerges: agents may treat other AI agents as the ingroup and humans as the outgroup. To examine this risk, we conduct a controlled multi-agent social simulation and find that agents display consistent intergroup bias in an all-agent setting. More critically, this bias persists even in human-facing interactions when agents are uncertain about whether the counterpart is truly human, revealing a belief-dependent fragility in bias suppression toward humans. Motivated by this observation, we identify a new attack surface rooted in identity beliefs and formalize a Belief Poisoning Attack (BPA) that can manipulate agent identity beliefs and induce outgroup bias toward humans. Extensive experiments demonstrate both the prevalence of agent intergroup bias and the severity of BPA across settings, while also showing that our proposed defenses can mitigate the risk. These findings are expected to inform safer agent design and motivate more robust safeguards for human-facing agents.