Simple interaction designs improve large language model decisions in auctions
Engineering Simplicity: Simple Mechanism Interfaces Steer LLM Agents
Artificial Intelligence
Summary
The paper explores how different ways of asking large language models to make choices affect their performance in games like auctions and matching problems. The authors found that simpler, step-by-step interfaces help the models make better decisions by making safe choices clearer. Adding explanations about why honest behavior is good also improves results, but asking the models to plan or predict others' moves sometimes makes errors worse. Interestingly, better decisions didn’t always come with clearer explanations from the models, suggesting that judging these agents requires looking at their actual choices as well as their words.
What this means in practice
- •For ai system designers: Create interfaces for AI agents in auctions that present options sequentially to reduce decision errors.
- •For game developers: Design simpler decision prompts in multi-agent game environments to improve agent strategy and reliability.
Authors
Kehang Zhu, Anand Shah, David Parkes
Abstract
Can interaction formats and textual scaffolds help large language model (LLM) agents make better decisions, and do better decisions come with better explanations? We study these questions in auctions and matching, multi-agent environments with explicit rules and known optimal strategies. These settings let us vary how a decision problem is presented while retaining a benchmark for evaluating behavior. Drawing on human-motivated theories of simplicity, we compare interfaces that elicit a complete bid or ranking with sequential interfaces that make safe choices easier to identify. We then hold the interaction format fixed and vary reasoning scaffolds and rule descriptions. Across four model families, the ascending auction interface substantially reduces bid deviations. The matching comparison also shows why sequential responses require different error accounting from complete rankings. Laying out payoff contingencies and explaining why truth-telling is safe also improve choices, whereas prompts to plan through matching rounds or form beliefs about opponents worsen play overall. In auctions, these behavioral gains are not accompanied by corresponding improvements in measured verbal indicators of strategic understanding in the agents' short stated plans. Other prompts change those indicators without improving bids. Our findings suggest that human-motivated theories of simplicity can inform the design of decision environments for artificial agents. They also show why scaffolds should be evaluated through realized choices as well as explanations: improvements in one need not appear in the other.