[Paper Review] "Guinea Pig Trials" Utilizing GPT: A Novel Smart Agent-Based Modeling Approach for Studying Firm Competition and Collusion

Xu Han, Zengqing Wu+1 more|arXiv (Cornell University)|Aug 21, 2023

Auction Theory and ApplicationsDecision Sciences3 citations

TL;DR

This paper introduces Smart Agent-Based Modeling (SABM), a novel framework using GPT-4-powered smart agents to simulate firm competition and collusion. Unlike traditional agent-based models, SABM agents exhibit human-like strategic reasoning and communication, enabling realistic simulation of market dynamics. The key finding is that communication accelerates collusion formation and leads to prices near cartel levels, while no communication results in tacit collusion at prices above Bertrand but below monopoly levels.

ABSTRACT

Firm competition and collusion involve complex dynamics, particularly when considering communication among firms. Such issues can be modeled as problems of complex systems, traditionally approached through experiments involving human subjects or agent-based modeling methods. We propose an innovative framework called Smart Agent-Based Modeling (SABM), wherein smart agents, supported by GPT-4 technologies, represent firms, and interact with one another. We conducted a controlled experiment to study firm price competition and collusion behaviors under various conditions. SABM is more cost-effective and flexible compared to conducting experiments with human subjects. Smart agents possess an extensive knowledge base for decision-making and exhibit human-like strategic abilities, surpassing traditional ABM agents. Furthermore, smart agents can simulate human conversation and be personalized, making them ideal for studying complex situations involving communication. Our results demonstrate that, in the absence of communication, smart agents consistently reach tacit collusion, leading to prices converging at levels higher than the Bertrand equilibrium price but lower than monopoly or cartel prices. When communication is allowed, smart agents achieve a higher-level collusion with prices close to cartel prices. Collusion forms more quickly with communication, while price convergence is smoother without it. These results indicate that communication enhances trust between firms, encouraging frequent small price deviations to explore opportunities for a higher-level win-win situation and reducing the likelihood of triggering a price war. We also assigned different personas to firms to analyze behavioral differences and tested variant models under diverse market structures. The findings showcase the effectiveness and robustness of SABM and provide intriguing insights into competition and collusion.

Motivation & Objective

• To develop a scalable, cost-effective alternative to human subject experiments for studying firm competition and collusion.
• To address limitations of traditional agent-based models by endowing agents with advanced reasoning and communication capabilities using large language models.
• To investigate how communication influences collusion formation, price convergence, and strategic behavior in oligopolistic markets.
• To examine the impact of firm personas and market structures on collusion dynamics using simulated agents.

Proposed method

• SABM employs GPT-4 as the cognitive engine for each firm agent, enabling complex decision-making and natural language interaction.
• Agents are endowed with a knowledge base and strategic reasoning to simulate real-world firm behavior in pricing and collusion decisions.
• The model simulates repeated interactions between firms under varying conditions: with and without communication, and under different market structures.
• Agents engage in price-setting games, with outcomes recorded to analyze convergence patterns and collusion stability.
• Personas are assigned to agents to simulate heterogeneity in firm behavior, such as aggressive vs. cooperative types.
• The framework is validated through controlled simulations, comparing outcomes across communication and non-communication scenarios.

Experimental results

Research questions

• RQ1 How does the presence of communication between firms affect the speed and stability of collusion formation in a competitive market?
• RQ2 To what extent do GPT-powered smart agents replicate human-like strategic behavior in oligopoly pricing games?
• RQ3 What is the impact of firm-specific personas on collusion dynamics and pricing outcomes in simulated markets?
• RQ4 How do different market structures influence the emergence and sustainability of collusion in the SABM framework?
• RQ5 Can SABM replicate known economic equilibria such as Bertrand and cartel pricing under controlled conditions?

Key findings

• In the absence of communication, smart agents consistently achieve tacit collusion, with prices converging to levels above the Bertrand equilibrium but below monopoly prices.
• When communication is enabled, agents form a higher-level collusion, achieving prices close to cartel equilibrium levels.
• Communication significantly accelerates the formation of collusion, reducing the time to price convergence compared to non-communicating scenarios.
• With communication, agents exhibit more frequent small price deviations, suggesting exploration of higher-win-win outcomes and reduced risk of price wars.
• The introduction of distinct firm personas leads to observable behavioral differences, with more cooperative agents contributing to faster and more stable collusion.
• The SABM framework demonstrates robustness across diverse market structures, confirming its viability as a scalable alternative to human subject experiments.