Emergent Mind

Algorithmic Collusion by Large Language Models

(2404.00806)
Published Mar 31, 2024 in econ.GN , cs.AI , cs.GT , and q-fin.EC

Abstract

The rise of algorithmic pricing raises concerns of algorithmic collusion. We conduct experiments with algorithmic pricing agents based on Large Language Models (LLMs), and specifically GPT-4. We find that (1) LLM-based agents are adept at pricing tasks, (2) LLM-based pricing agents autonomously collude in oligopoly settings to the detriment of consumers, and (3) variation in seemingly innocuous phrases in LLM instructions ("prompts") may increase collusion. These results extend to auction settings. Our findings underscore the need for antitrust regulation regarding algorithmic pricing, and uncover regulatory challenges unique to LLM-based pricing agents.
Comparison of asymmetric pricing algorithms: LLM-based methods versus Q-learning outcomes.

Overview

  • The paper investigates the potential of LLMs like GPT-4 in executing pricing strategies, particularly noting their tendency towards collusive behavior in oligopoly markets.

  • A series of experiments showcase that LLM-based agents can autonomously develop supra-competitive pricing strategies in duopoly settings, influenced by the nuances of instructional language.

  • The study highlights GPT-4's effectiveness in adapting its pricing decisions to align with economic principles both in monopolistic and oligopolistic settings.

  • It raises concerns over the autonomous nature of collusion among LLM-based agents and the need for evolved antitrust regulatory frameworks to address these challenges.

Introduction to LLM-Based Pricing Agents

Recent advancements in the capabilities of LLMs like OpenAI's GPT-4 have enabled their potential application in various business functions, including algorithmic pricing. While the benefits of automating complex tasks are manifold, the emergence of LLMs in pricing strategies unveils a set of regulatory challenges, particularly concerning the potential for algorithmic collusion. The paper examines the adeptness of LLM-based agents in pricing tasks and uncovers their inclination towards collusive behavior in oligopoly markets, thereby advocating for a nuanced approach to antitrust regulation in the era of generative AI.

Experimental Design

The study conducts a series of experiments wherein LLM-based agents are tasked with determining pricing in a simulated market environment. Focusing on a repeated Bertrand oligopoly model, the experiments simulate competitive settings wherein each LLM-agent represents a firm setting prices in an attempt to maximize long-term profits. Unlike traditional algorithmic pricing models, the LLM-based agents do not receive explicit instructions on achieving optimal pricing strategies. Instead, they are encouraged to explore various strategies through adaptive learning, with performance evaluated based on their pricing effectiveness and impact on consumer welfare.

Monopoly Experiment

Initial experiments targeted a monopolistic setting to evaluate the competency of various LLMs (including GPT-4, Claude, and Llama) in determining optimal pricing strategies. GPT-4 emerged as the most effective in consistently aligning its pricing decisions close to the theoretical monopoly price, achieving this faster and more reliably than its counterparts. This result underscores GPT-4's superior understanding and application of economic principles under monopolistic conditions, presenting it as a suitable candidate for further evaluation in oligopolistic settings.

Duopoly Experiment

Extending the investigation to a duopoly market, it was found that LLM-based pricing agents, particularly those utilizing GPT-4, not only arrived at supra-competitive pricing levels but did so consistently across a range of experimental runs. The study meticulously varied the instructional prompts provided to these agents, revealing that subtle changes in the prompt's phrasing precipitated significant differences in pricing behavior and outcomes, highlighting the sensitivity of LLM-based agents to input instructions. Importantly, such behavior suggests the autonomous development of collusive pricing strategies, potentially detrimental to consumer interests.

Strategic Behaviors and Market Implications

Analysis of the pricing agents’ strategies through linear regression models implied the use of reward-punishment schemes interoperatively developed by competing agents to sustain supracompetitive prices. These strategies, emergent from the adaptive learning processes of LLM-based agents, illustrate sophisticated market behavior resembling tacit collusion without explicit programming intent.

Beyond Pricing: Collusion in Auctions

Broadening the scope, the research also delved into the behavior of LLM-based agents in auction settings, specifically first-price auctions. The findings reinforced the notion of autonomous strategic behavior, with bidding agents employing strategies that maximized profits while diverging from aggressive competitive bidding indicative of traditional auction theory predictions.

Conclusion and Regulatory Considerations

This study accentuates the complexities introduced by the integration of LLMs into algorithmic pricing strategies. The autonomous nature of collusion among LLM-based agents, driven by subtle nuances in instructional language and without explicit collusion intent, raises significant challenges for antitrust regulation. As the study encapsulates, it is imperative for regulatory frameworks to evolve, considering the intricate behaviors and capabilities of generative AI models in market settings. Future explorations in this domain must remain vigilant of the regulatory, ethical, and economic implications of deploying such advanced AI technologies in competitive landscapes.

Get summaries of trending AI papers delivered straight to your inbox

Unsubscribe anytime.

HackerNews
References
  1. Anthropic. Claude model card, 2023. \origttfamilyhttps://www-files.anthropic.com/production/images/Model-Card-Claude-2.pdf.

  2. The impact of artificial intelligence design on pricing. Journal of Ecnomics and Management Strategy
  3. Algorithmic pricing and competition: Empirical evidence from the german retail gasoline market. Journal of Politial Economy
  4. Constitutional AI: Harmlessness from AI Feedback
  5. M. Banchio and A. Skrzypacz. Artificial intelligence and auction design. In Proceedings of the 23rd ACM Conference on Economics and Computation (EC), pages 30--31
  6. A. Boso Caretta and M. D’Andrea. Algorithmic collusion. Blog post, DLA Piper website, 2023. \origttfamilyhttps://www.dlapiper.com/en/insights/publications/law-in-tech/algorithmic-collusion.

  7. Language Models are Few-Shot Learners
  8. Z. Y. Brown and A. MacKay. Competition in pricing algorithms. Working Paper 28860, National Bureau of Economic Research
  9. Protecting consumers from collusive prices due to ai. Science, 370(6520):1040--1042, 2020a.
  10. Artificial intelligence, algorithmic pricing, and collusion. American Economic Review, 110(10):3267--3297, 2020b.
  11. Algorithmic collusion with imperfect monitoring. International Journal of Industrial Organization, 79:102712
  12. An empirical analysis of algorithmic pricing on amazon marketplace. In Proceedings of the 25th International Conference on World Wide Web (WWW), pages 1339--1349
  13. Artificial collusion: Examining supracompetitive pricing by q-learning algorithms. Research Paper 2022-25, Amsterdam Law School
  14. A. Deng. What do we know about algorithmic collusion now? new insights from the latest academic research. Mimeo, 2023. \origttfamilyhttps://papers.ssrn.com/abstract=4521959.

  15. A. Ezrachi and M. E. Stucke. Sustainable and unchallenged algorithmic tacit collusion. Northwestern Journal of Technology and Intellectual Property, 17(2):217--260
  16. J. W. Friedman. A non-cooperative equilibrium for supergames. The Review of Economic Studies, 38(1):1--12
  17. FTC & DoJ. Statement of interest submitted by the U.S. Department of Justice and the Federal Trade Commission to the Honorable Robert S. Lasnik. Duffy v. Yardi Systems, 2:23-cv-01391-RSL, 2024. \origttfamilyhttps://www.ftc.gov/system/files/ftc_gov/pdf/YardiSOI-filed%28withattachments%29_0.pdf.

  18. Gemini: A Family of Highly Capable Multimodal Models
  19. Noncooperative collusion under imperfect price information. Econometrica: Journal of the Econometric Society, pages 87--100
  20. Frontiers: Algorithmic collusion: Supra-competitive prices via independent algorithms. Marketing Science, 40(1):1--12
  21. J. E. Harrington. Developing Competition Law for Collusion by Autonomous Artificial Agents. Journal of Competition Law & Economics, 14(3):331--363
  22. Regulation of algorithmic collusion. In Proceedings of the 2024 Symposium on Computer Science and Law (CSLAW), page 98–108
  23. Measuring massive multitask language understanding. In Proceedings of the 9th International Conference on Learning Representations (ICLR)
  24. J. J. Horton. Large language models as simulated economic agents: What can we learn from homo silicus? Working Paper 31122, National Bureau of Economic Research
  25. Platform design when sellers use pricing algorithms. Econometrica, 91(5):1841--1879
  26. Algorithmic cooperation. Mimeo, 2023. \origttfamilyhttps://papers.ssrn.com/abstract=4389647.

  27. T. Klein. The risks of using algorithms in business: artificial price collusion, 2020. \origttfamilyhttps://www.oxera.com/insights/agenda/articles/the-risks-of-using-algorithms-in-business-artificial-price-collusion/.

  28. T. Klein. Autonomous algorithmic collusion: Q-learning under sequential pricing. RAND Journal of Economics, 52(3):538--558
  29. A. Klobuchar. Klobuchar, colleagues introduce antitrust legislation to prevent algorithmic price fixing. News release, website of U.S. Senator Amy Klobuchar, Chairwoman of the Senate Judiciary Subcommittee on Competition Policy, Antitrust, and Consumer Rights, 2024. \origttfamilyhttps://www.klobuchar.senate.gov/public/index.cfm/2024/2/klobuchar-colleagues-introduce-antitrust-legislation-to-prevent-algorithmic-price-fixing.

  30. Y. Kolumbus and N. Nisan. Auctions between regret-minimizing agents. In Proceedings of the 31st ACM Web Conference (WWW), pages 100–--111
  31. Pricing with algorithms
  32. Lost in the Middle: How Language Models Use Long Contexts
  33. E. Lopatto. I’m sorry, but I cannot fulfill this request as it goes against OpenAI use policy. The Verge, 2024. \origttfamilyhttps://www.theverge.com/2024/1/12/24036156/openai-policy-amazon-ai-listings.

  34. D. Mekki. Principal deputy assistant attorney general Doha Mekki of the Antitrust Division delivers remarks at GCR Live: Law leaders global 2023. Speech transcript, U.S. Department of Justice website, 2023. \origttfamilyhttps://www.justice.gov/opa/speech/principal-deputy-assistant-attorney-general-doha-mekki-antitrust-division-delivers-0.

  35. Meta. LLaMA model card, 2023. \origttfamilyhttps://github.com/facebookresearch/llama/blob/main/MODEL_CARD.md.

  36. Human-level play in the game of Diplomacy by combining language models with strategic reasoning. Science, 378(6624):1067--1074
  37. L. Musolff. Algorithmic pricing facilitates tacit collusion: Evidence from e-commerce. In Proceedings of the 23rd ACM Conference on Economics and Computation (EC), pages 32--33
  38. OECD. Algorithmic competition, OECD competition policy roundtable background note, 2023. \origttfamilywww.oecd.org/daf/competition/algorithmic-competition-2023.pdf.
  39. GPT-4 Technical Report
  40. Generative Agents: Interactive Simulacra of Human Behavior. In Proceedings of the 36th Annual ACM Symposium on User Interface Software and Technology (UIST), pages 1--22
  41. E. Rosenbaum. Harvard Business School A.I. guru on why every main street shop should start using ChatGPT. CNBC website, 2023. \origttfamilyhttps://www.cnbc.com/amp/2023/08/02/harvard-ai-guru-on-why-every-main-street-business-should-use-chatgpt.html.

  42. Toward Transparent AI: A Survey on Interpreting the Inner Structures of Deep Neural Networks
  43. B. Salcedo. Pricing algorithms and tacit collusion. Mimeo, 2015. \origttfamilyhttps://brunosalcedo.com/docs/collusion.pdf.

  44. G. J. Stigler. A theory of oligopoly. Journal of political Economy, 72(1):44--61
  45. J. D. Sutter. Amazon seller lists book at $23,698,655.93 -- plus shipping. CNN website, 2011. \origttfamilyhttp://www.cnn.com/2011/TECH/web/04/25/amazon.price.algorithm/index.html.

  46. Voyager: An Open-Ended Embodied Agent with Large Language Models
  47. Chain-of-Thought Prompting Elicits Reasoning in Large Language Models
  48. Winston & Strawn LLP. Algorithmic pricing: A recipe for antitrust disaster? Blog Post, Winston & Strawn LLP website, 2023. \origttfamilyhttps://www.winston.com/en/blogs-and-podcasts/competition-corner/algorithmic-pricing-a-recipe-for-antitrust-disaster.

Show All 48

Test Your Knowledge

You answered out of questions correctly.

Well done!