Language model agents show limited strategic flexibility in group risk games
Cooperation Without Adaptation: LLM Agents in a Collective-Risk Dilemma
Computer Science and Game TheoryArtificial IntelligenceComputation and LanguageMultiagent Systems
Summary
This paper looks at how language model agents behave when making shared decisions that affect everyone’s outcome in a risky situation. The authors found that these agents mostly stick to fixed strategies and do not adjust their contributions based on changing risks or what other players do. Even when contributing is not beneficial, some agents still contribute at zero risk. How agents respond to incentives depends on their group and the mix of agents they interact with. This means just seeing cooperation doesn’t mean these agents adapt smartly to others or changing conditions.
What this means in practice
- •For multi-agent system engineers: Design tests that check if language model agents adapt to changing partner behavior and incentives in collaborative risk settings.
- •For ai governance teams: Evaluate and monitor deployed AI agents for responsiveness to group incentives and partner actions to avoid unexpected group failures.
Authors
Trung-Kiet Huynh, Dao-Sy Duy-Minh, Chi-Nguyen Tran, Phu-Hoa Pham, Phu-Quy Nguyen-Lam, Le Hong Trang, The Anh Han
Abstract
Language-model agents are increasingly deployed to act on behalf of people in settings where individual decisions shape shared outcomes. We examine whether five production models adapt their contributions to changing risks and partner behaviour, or instead follow a relatively fixed strategy. In a collective-risk dilemma game (CRG), six players contribute to a shared pool over ten rounds; if the pool falls short of its target, a catastrophe may destroy their remaining resources. Four models contributed approximately an equal share at every positive risk level, and four of the five also contributed when risk was zero, even though contributing could not improve their payoff. Explicitly stating that only the agent's own payoff mattered reduced zero-risk contributions for most models, while a more neutral prompt did not. Several models correctly identified which action offered the higher expected payoff at different risk levels, yet their contributions changed little. Against fixed partners, no model consistently chose the payoff-maximising response, and in mixed-model groups, a single agent's shortfall in the final round could cause an otherwise successful group to fail. Payoff-based selection also depended on population composition: in a single six-agent group, it favoured the model that contributed least and reduced group success, whereas in larger, randomly mixed populations, it favoured an exact fair-share model at moderate and high risk. These findings show that cooperative self-play alone is not evidence of strategic adaptability. Consistent with a machine-behaviour approach, evaluations of agents in multi-agent systems should test how they respond to incentives, partner actions, and population composition.