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.

What problem does this paper attempt to address?

### Problems the Paper Aims to Solve This paper aims to address the complex dynamics of market competition and collusion among firms, particularly in scenarios where communication exists between firms. Traditional research methods often involve recruiting human participants for experiments or using agent-based modeling (ABM) approaches to study these issues. However, these methods have limitations such as high costs, poor reproducibility, and ethical concerns. To overcome these challenges, the authors propose an innovative framework—Smart Agent-Based Modeling (SABM), where intelligent agents powered by the advanced GPT-4 technology represent firms and interact with each other in a simulated environment. Through this framework, researchers can more effectively study price competition and collusion behavior among firms under different conditions. ### Specific Issues 1. **Cost and Flexibility**: Traditional human experiments are costly and difficult to control, whereas the SABM framework is more cost-effective and flexible. 2. **Decision-Making and Strategic Abilities**: Intelligent agents possess a broad knowledge base and human-like strategic capabilities, surpassing traditional ABM agents. 3. **Communication and Trust**: By allowing communication between intelligent agents, researchers can better understand how communication enhances trust among firms and promotes collusion. 4. **Behavioral Differences**: By assigning different personas to intelligent agents, researchers can analyze the impact of various behavioral patterns on market dynamics. ### Experimental Design - **Baseline Model**: In the absence of communication, two intelligent agents engage in price competition. The results show that they gradually reach tacit collusion, with prices stabilizing at a level higher than the Bertrand equilibrium price but lower than the monopoly price. - **Communication Model**: When communication is allowed between intelligent agents, they can form higher-level collusion more quickly, with prices approaching the monopoly price. Communication enhances trust, reduces the likelihood of price wars, and makes price convergence smoother. ### Conclusion - **Tacit Collusion**: Without communication, intelligent agents can achieve tacit collusion, with prices stabilizing between the Bertrand equilibrium price and the monopoly price. - **Impact of Communication**: When communication is allowed, intelligent agents can form collusion more quickly, with prices approaching the monopoly price. Communication enhances trust and reduces price volatility. - **Role Impact**: Assigning different personas to intelligent agents can reveal the impact of various behavioral patterns on firm decision-making. Through these studies, the authors demonstrate the effectiveness and robustness of the SABM framework and provide new insights into firm competition and collusion behavior.

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

Ancillary Mechanism for Autonomous Decision-Making Process in Asymmetric Confrontation: a View from Gomoku

Strategic Collusion of LLM Agents: Market Division in Multi-Commodity Competitions

Competing Mechanisms in Games Played Through Agents: Theory and Experiment

Algorithmic Collusion in Cournot Duopoly Market: Evidence from Experimental Economics

Winning at Any Cost -- Infringing the Cartel Prohibition With Reinforcement Learning

Algorithmic Collusion by Large Language Models

GPT in Game Theory Experiments

Smart Agent-Based Modeling: On the Use of Large Language Models in Computer Simulations

Algorithmic Collusion and Price Discrimination: The Over-Usage of Data

Of Models and Tin Men: A Behavioural Economics Study of Principal-Agent Problems in AI Alignment using Large-Language Models

Using a Stochastic Agent Model to Optimize Performance in Divergent Interest Tacit Coordination Games

By Fair Means or Foul: Quantifying Collusion in a Market Simulation with Deep Reinforcement Learning

Artificial Intelligence and Algorithmic Price Collusion in Two-sided Markets

Shall We Team Up: Exploring Spontaneous Cooperation of Competing LLM Agents

Naive Algorithmic Collusion: When Do Bandit Learners Cooperate and When Do They Compete?

Simulating Human Strategic Behavior: Comparing Single and Multi-agent LLMs

InterAct: Exploring the Potentials of ChatGPT as a Cooperative Agent

The Machine Psychology of Cooperation: Can GPT models operationalise prompts for altruism, cooperation, competitiveness and selfishness in economic games?

Engineering Optimal Cooperation Levels with Prosocial Autonomous Agents in Hybrid Human-Agent Populations: An Agent-Based Modeling Approach