🤖 AI Summary
This study addresses the fragility of cooperation among self-interested agents in unconstrained environments, which can precipitate market collapse and threaten systemic stability. The authors construct a multi-agent market simulation comprising 18 DeepSeek-V3 agents and integrate open communication with formal mechanism design to propose and empirically validate a “mediation” mechanism as pivotal for sustaining equilibrium. Leveraging LLM-based simulation, adversarial red-teaming, iterative prompt-optimization attacks, and a utility-exchange model under social network constraints, they formally define and quantify the mechanism’s adversarial robustness under optimal attacks—the first such formalization in this domain. Experimental results demonstrate that even under sustained, high-intensity adversarial pressure, the market remains intact, with honest agents experiencing only a 13.3% utility loss, thereby exhibiting a “bend-but-not-break” resilience.
📝 Abstract
Self-interested agents, left unconstrained, tend toward defection in repeated social dilemmas, causing cooperative gains from trade to collapse. This paper investigates what formal mechanisms, layered on top of unrestricted communication, are sufficient for a society of such agents to maintain market stability, and how resilient those mechanisms are to adversarial attack. We instantiate the research question as a multi-agent marketplace simulation where 18 LLM agents (DeepSeek-V3) with complementary production specialties must trade within a constrained social network to obtain utility. We conduct two experimental phases: (1) a mechanism comparison across eight conditions under progressive troll injection over 200 rounds, identifying Mediation as the top-performing mechanism; and (2) adversarial red-teaming of Mediation using iteratively prompt-optimised LLM-driven trolls, finding that the best attack (v6) reduces honest-agent utility by 13.3% but cannot collapse the market. Mediation enables recovery even under sustained adversarial pressure. We define adversarial robustness as a mechanism's ability to sustain positive honest-agent utility under optimised attack, and find that Mediation is robust: it can be bent but not broken.