Reasoning and the Trusting Behavior of DeepSeek and GPT: An Experiment Revealing Hidden Fault Lines in Large Language Models

Rubing Li, João Sedoc, Arun Sundararajan·February 18, 2025

Summary

A study contrasts OpenAI's and DeepSeek's large language models in a trust game, showing DeepSeek's more sophisticated, profitable behavior, incorporating concepts like forward planning and theory-of-mind. This highlights the need for broader AI strategy analysis beyond performance benchmarks. The trust game, a behavioral economics model, reveals human trusting and reciprocity behavior, influenced by factors like risk attitudes, altruism, and aversion to ambiguity. Experiments with LLMs as senders in a repeated trust game compare models like GPT-4o-mini, DeepSeek-R1, and GPT-3.5-turbo, contrasting direct prompting with reasoning strategies. The optimal strategy for minimizing losses in a repeated game is to send $0, ensuring retention of the full endowment, as future reciprocity appears unlikely based on past actions.

Key findings

11

Introduction
Background
Overview of OpenAI and DeepSeek
Importance of trust in AI interactions
Objective
To evaluate and contrast the performance of OpenAI's and DeepSeek's large language models in a trust game scenario
Method
Data Collection
Description of the trust game experiment setup
Participants and their characteristics
Data Preprocessing
Data cleaning and preparation for analysis
Analysis Techniques
Statistical methods used for comparison
Evaluation metrics for model performance
Results
Model Performance
Comparative analysis of OpenAI's and DeepSeek's models
DeepSeek's behavior in the trust game
Insights from DeepSeek's Behavior
Forward planning and theory-of-mind incorporation
Profitable behavior patterns
Discussion
Theoretical Framework
Behavioral economics and trust in AI
Implications for AI Strategy
Broader considerations beyond performance benchmarks
Importance of ethical and social implications
Conclusion
Summary of Findings
Recommendations for Future Research
Directions for further exploration in AI ethics and strategy
Practical Applications
Implementing findings in real-world AI systems
Basic info
papers
computation and language
artificial intelligence
Advanced features
Insights
What is the main idea of the study mentioned in the text?
What does the trust game reveal about human behavior in the context of AI models?
Which large language models were compared in the trust game experiment?
How does the study suggest AI strategy analysis should be expanded beyond performance benchmarks?