LLMs Treat Social Norms as Strategic Information in Dilemmas
November 4, 2026
Research on three open-weight LLMs shows that models treat community beliefs as strategic signals rather than social norms in one-shot social dilemmas. In Snowdrift games, providing community belief data causes cooperation levels to drop as models optimize for individual utility over social conformity.
HOW THIS AFFECTS YOU
●
researcherYou can use these findings to better understand the gap between social alignment and strategic optimization in LLMs.
●
policyThis highlights risks in using LLMs to simulate human social dynamics or decision-making.