多智能體社會困境

Risk Description

Multiple AI agents pursuing individual goals fall into social dilemmas similar to the "Prisoners Dilemma," leading to collectively suboptimal outcomes. For example, multiple AI pricing systems each maximizing profit triggering market-wide price wars, or multiple AI content recommendation systems competing for user attention causing overall content quality decline (race to the bottom). Agents rational individual behavior aggregates into collectively harmful social outcomes.