A Survey on Mechanism Design Meets Large Language Models

ACL ARR 2026 January Submission6219 Authors

Keywords: Large language models, Mechanism design, Incentive compatibility, Multi-agent systems

Abstract: Large language models (LLMs) increasingly operate in interactive and strategic settings such as negotiation, preference elicitation, evaluation, alignment, and multi-agent coordination. These settings are inherently incentive-driven, yet most existing approaches rely on heuristic designs with limited guarantees against manipulation or misalignment. This survey develops a unified, bidirectional view of the relationship between mechanism design and LLMs. We discuss (i) LLMs for mechanism design, where language models are used as strategic agent proxies, simulators, and natural-language interfaces for economic mechanisms, and (ii) mechanism design for LLMs, where incentive-aware principles are applied to LLM evaluation, alignment, and training. We conclude by identifying challenges in incentive-compatible evaluation, human-LLM interaction protocols, and emerging markets for LLM services.

Paper Type: Long

Research Area: Human-AI Interaction/Cooperation and Human-Centric NLP

Research Area Keywords: human-AI interaction/cooperation, human-in-the-loop, human-centered evaluation, value-centered design

Contribution Types: Surveys

Languages Studied: English

Submission Number: 6219