Algorithmic Collusion by Large Language Models

2024/03/31 by Sara Fish, Yannai A. Gonczarowski, Fish, Sara +3 · 7 voices · 14 citations
Computer Science · #Machine Learning and Algorithms #Natural Language Processing Techniques #Topic Modeling #cs.AI #cs.GT #econ.GN

paper · pdf · doi:10.48550/arxiv.2404.00806

openalex publication_date 2024/03/31 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

We conduct experiments with algorithmic pricing agents based on Large Language Models (LLMs). In oligopoly settings, LLM-based pricing agents quickly and autonomously reach supracompetitive prices and profits. Variation in seemingly innocuous phrases in LLM instructions ("prompts") substantially influence the degree of supracompetitive pricing. We develop novel techniques for behavioral analysis of LLMs and use them to uncover price-war concerns as a contributing factor. Our results extend to auction settings. Our findings uncover unique challenges to any future regulation of LLM-based pricing agents, and AI-based pricing agents more broadly.

Cited by

• Draining the Energy Commons: Self-Defeating Over-Appropriation as a Coordination Failure in Agentic LLM Collectives
• When Is Delegated Play Truthful? Within-Range Regret and the Trilemma of Aligned Delegation
• Markets are competitive if and only if P != NP
• Solipsistic Superintelligence is Unlikely to be Cooperative
• Beyond Single-Agent Safety: A Taxonomy of Risks in LLM-to-LLM Interactions
• Breaking Algorithmic Collusion in Human-AI Ecosystems
• Tacit Bidder-Side Collusion: Artificial Intelligence in Dynamic Auctions
• GenAI vs. Human Creators: Procurement Mechanism Design in Two-/Three-Layer Markets
• A Criminology of Machines
• Magentic Marketplace: An Open-Source Environment for Studying Agentic Markets
• Agentic markets
• Strategic Tradeoffs Between Humans and AI in Multi-Agent Bargaining
• Too Noisy to Collude? Algorithmic Collusion Under Laplacian Noise
• The Collaboration Paradox: Why Generative AI Requires Both Strategic Intelligence and Operational Stability in Supply Chain Management

Discussions

• AI is good at pricing, so when GPT-4 was asked to help merchants maximize profits - and it did exactly that by secretly coordinating with other AIs to keep prices high! So... aligned for whom? Mercha [bsky, 114 points, 1 comments]
• In other words: technological advancement might have made the early warnings of Calvano, Calzolari, Denicolò, Harrington, Pastorello, and others an imminent reality. 3/3 arxiv.org/abs/2404.00806 [bsky, 8 points, 0 comments]
• #AI #EconSky #DataScience arxiv.org/abs/2404.00806 An interesting title, but maybe not as big a worry as it presents! A statistical model operating alongside similar others can entrain (fireflies or m [bsky, 5 points, 1 comments]
• 〝By demonstrating algorithmic collusion by an AI tool that is increasingly adopted for various business purposes, our findings emphasize the critical need for regulation of pricing and bidding algorit [bsky, 3 points, 0 comments]
• Algorithmic Collusion by Large Language Models [hn, 3 points, 0 comments]
• GPT-4 ist sehr gut im Preise verhandeln. Fast zu gut – tendiert zu Preisabsprachen und Kartellbildung wo möglich. arxiv.org/abs/2404.00806 [bsky, 1 points, 0 comments]
• « Algorithmic Collusion by Large Language Models∗ » arxiv.org/pdf/2404.00806 [bsky, 0 points, 0 comments]

Related

• Learning from Synthetic Labs: Language Models as Auction Participants
• Strategic Collusion of LLM Agents: Market Division in Multi-Commodity Competitions
• Tacit Bidder-Side Collusion: Artificial Intelligence in Dynamic Auctions
• Evaluating LLM Agent Collusion in Double Auctions
• By Fair Means or Foul: Quantifying Collusion in a Market Simulation with Deep Reinforcement Learning