Engineering Simplicity: Simple Mechanism Interfaces Steer LLM Agents

2026-09-28

AI Summary

This study addresses the tendency of large language model (LLM) agents to deviate from optimal strategies in multi-agent decision-making scenarios such as auctions, where their behaviors frequently exhibit inconsistencies with their generated explanations. Drawing upon human-inspired simplicity theory, this work optimizes mechanism interfaces through prompt engineering techniques—including sequential auction interfaces, payoff-contingent layouts, and explicit rule explanations—to decouple behavioral improvements from explanatory enhancements. Experimental results demonstrate a significant reduction in bidding deviations. Furthermore, the findings reveal the necessity of independently evaluating LLMs’ actual choices and verbal justifications, thereby establishing an effective paradigm for optimizing decision-making performance in multi-agent environments.

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.

Problem

• LLM agents
• interaction formats
• textual scaffolds
• decision making
• mechanism design

Innovation

• LLM Agents
• Mechanism Design
• Interaction Interfaces
• Reasoning Scaffolds
• Behavioral Evaluation