Chess pieces representing strategy and AI collaboration.

Can AI Chatbots Like ChatGPT Help Us Understand Human Cooperation?

Lena Kashyap in Tech & Innovation December 2025 • 4 min read.

"Exploring the Potential of Large Language Models in Economic Games"

Have you ever wondered why people cooperate, even when it seems like they should act in their own self-interest? Economists and psychologists have long studied this question using "economic games"—simulations where people make choices about how to share resources, cooperate, or compete. Now, researchers are turning to a new tool: large language models (LLMs) like ChatGPT.

These AI models, trained on vast amounts of text data, are surprisingly good at mimicking human-like conversation and behavior. But can they also simulate the complex decision-making that underlies cooperation and altruism? A recent study delves into this question, exploring whether ChatGPT can "operationalize" prompts for different social behaviors in classic economic games.

The study investigates how well ChatGPT can translate natural language descriptions of cooperative, competitive, altruistic, and self-interested behavior into appropriate actions within the context of economic games, such as the Prisoner's Dilemma and the Dictator Game. The goal is to see if these AI models can capture the nuances of human decision-making in social situations.

Testing AI's Capacity for Cooperation

Chess pieces representing strategy and AI collaboration.

The researchers designed a series of experiments using ChatGPT-3.5, an earlier version of the well-known AI chatbot. They used prompts—textual instructions—to create simulated individuals ("simulacra") with different personalities: cooperative, competitive, altruistic, and selfish. These simulacra then played two classic economic games:

The Prisoner's Dilemma: This game explores cooperation and competition. Two players can choose to cooperate or defect. If both cooperate, they both receive a reward. If one defects and the other cooperates, the defector gets a larger reward, and the cooperator gets nothing. If both defect, they both receive a smaller punishment.

The Dictator Game: This game tests altruism. One player (the dictator) decides how to split a sum of money between themselves and another player (the recipient). The recipient has no say in the decision.

The researchers manipulated the prompts to create different versions of the simulacra. For example, a "cooperative" simulacrum might be told to value teamwork and seek the best outcome for all parties, while a "selfish" simulacrum might be described as a cunning strategist who prioritizes their own well-being above all else. The AI's choices in these games were then analyzed to see if they aligned with their assigned personalities.

What Does This Mean for the Future?

This research suggests that AI models like ChatGPT have the potential to help us understand the complexities of human social behavior. By creating AI agents that mimic different personality traits and observing how they interact in economic games, we can gain new insights into the factors that drive cooperation, competition, and altruism. This could have implications for designing more effective strategies for promoting cooperation in real-world scenarios, from business negotiations to public policy.

About this Article -

This article was crafted using a human-AI hybrid and collaborative approach. AI assisted our team with initial drafting, research insights, identifying key questions, and image generation. Our human editors guided topic selection, defined the angle, structured the content, ensured factual accuracy and relevance, refined the tone, and conducted thorough editing to deliver helpful, high-quality information.See our About page for more information.

This article is based on research published under:

DOI-LINK: https://doi.org/10.48550/arXiv.2305.0797,

Title: The Machine Psychology Of Cooperation: Can Gpt Models Operationalise Prompts For Altruism, Cooperation, Competitiveness And Selfishness In Economic Games?

Subject: cs.gt cs.ai cs.cy econ.gn q-fin.ec

Authors: Steve Phelps, Yvan I. Russell

Published: 13-05-2023

Everything You Need To Know

What is the primary goal of using AI chatbots like ChatGPT in the study of human cooperation?

The main objective is to leverage AI models, specifically ChatGPT, to gain a deeper understanding of human behavior in economic situations. Researchers use the AI to simulate various social behaviors, such as cooperation, competition, altruism, and self-interest, within the framework of economic games. By analyzing how these AI models respond to different scenarios, researchers aim to identify the underlying factors that influence human decision-making in real-world settings.

How do economic games like the Prisoner's Dilemma and the Dictator Game help researchers study AI behavior and human behavior?

Economic games are used to simulate decision-making scenarios. The Prisoner's Dilemma tests cooperation and competition where two players choose to cooperate or defect. The outcomes depend on both players' choices. The Dictator Game, on the other hand, explores altruism, with one player deciding how to split resources. ChatGPT-3.5 is used to create "simulacra" – AI agents with different personalities (cooperative, competitive, altruistic, selfish) and then observe their choices in these games. By analyzing the AI's choices, researchers can gain insights into how different personality traits influence decision-making in social situations. This helps correlate AI behavior with potential human behavior in similar scenarios.

Can you explain the role of "simulacra" and prompts in the context of using ChatGPT to study human behavior?

In this research, "simulacra" refer to simulated individuals or AI agents created within ChatGPT-3.5. These are programmed using textual instructions called "prompts" to exhibit specific personality traits such as cooperative, competitive, altruistic, or selfish behavior. Prompts are crucial because they define the AI's "personality" and guide its actions within economic games. For example, a prompt might instruct a simulacrum to prioritize teamwork, leading to cooperative actions in the Prisoner's Dilemma. The AI's responses and choices are analyzed to understand the impact of these different personalities on decision-making in simulated economic scenarios.

What are the potential implications of using AI models like ChatGPT for understanding human cooperation, competition, and altruism?

The research suggests that AI models such as ChatGPT have the potential to offer fresh insights into human social behavior. By creating AI agents with varied personality traits and observing their interactions in economic games, researchers can identify the factors that drive cooperation, competition, and altruism. These findings could influence real-world strategies, like improving negotiation tactics in business and enhancing public policy to encourage cooperation. Moreover, this approach could lead to a deeper understanding of how to design environments that promote positive social outcomes across different contexts.

How does the study described contribute to the field of economics and social sciences?

This study contributes to the field by introducing a new method for exploring human behavior using AI. By employing LLMs like ChatGPT, researchers can model different social behaviors and analyze the decision-making processes in controlled environments. This approach helps in operationalizing abstract concepts such as cooperation and altruism, by translating them into specific actions within economic games such as the Prisoner's Dilemma and the Dictator Game. Ultimately, the study offers new perspectives on the complexities of social interactions and provides tools that can be applied to create more effective strategies for fostering cooperation in practical situations, spanning from corporate negotiations to designing public policies.