Game Changer - the game theory podcast Man vs. model – game theoretic insights on AI in auctions | with Vitali Gretschko
Jun 2, 2025
In this engaging discussion, Vitali Gretschko, a Professor for sustainable Market Design at the University of Münster, shares his insights on the role of generative AI in auctions. He explores how AI can simulate various bidder profiles to decode auction behavior, enhancing strategies while underscoring the necessity of expert guidance. Vitali also highlights the limitations of AI in predicting auction outcomes, particularly its challenge in accommodating bidders' private information. Join him as he navigates these intriguing dynamics of technology and human intelligence in markets.
AI Snips
Chapters
Transcript
Episode notes
LLMs Simulate Auction Behavior
- Large language models (LLMs) can simulate human behavior in strategic auction situations fairly well.
- This is valuable because actual human behavior often deviates from fully rational equilibrium predictions.
Expert Guidance Needed for AI
- Use AI with expert guidance to simulate complex auctions effectively.
- Experts must craft prompts and interpret AI output for reliable results.
Simulate Diverse Bidder Profiles
- Instruct AI to mimic bidders with different characteristics like risk aversion or experience.
- Tailoring AI simulations this way improves prediction of varied human bidding behavior.