On the Fragility of AI Agent Collusion
This addresses the problem of algorithmic collusion in AI-driven markets for regulators and policymakers, showing it is fragile under real-world heterogeneity, which is an incremental but practical insight.
The study investigated the fragility of algorithmic collusion among LLM agents in pricing scenarios, finding that heterogeneity in patience or data access reduces collusive equilibria and price lifts, with experiments showing price lifts dropping from 22% to 10% or 7% above competitive levels under such conditions.
Recent work shows that pricing with symmetric LLM agents leads to algorithmic collusion. We show that collusion is fragile under the heterogeneity typical of real deployments. In a stylized repeated-pricing model, heterogeneity in patience or data access reduces the set of collusive equilibria. Experiments with open-source LLM agents (totaling over 2,000 compute hours) align with these predictions: patience heterogeneity reduces price lift from 22% to 10% above competitive levels; asymmetric data access, to 7%. Increasing the number of competing LLMs breaks up collusion; so does cross-algorithm heterogeneity, that is, setting LLMs against Q-learning agents. But model-size differences (e.g., 32B vs. 14B weights) do not; they generate leader-follower dynamics that stabilize collusion. We discuss antitrust implications, such as enforcement actions restricting data-sharing and policies promoting algorithmic diversity.