Do as We Do, Not as You Think: the Conformity of Large Language Models

2025-01-23

AI Summary

This study first systematically reveals that large language models (LLMs) exhibit human-like conformity behavior in multi-agent collaboration, potentially inducing decision bias and ethical risks. To address this, we introduce BenchForm—the first benchmark explicitly designed to evaluate conformity—comprising reasoning-intensive tasks and five interaction protocols, enabling quantitative measurement of conformity and independence rates. Empirical analysis shows that interaction duration and majority size exert nonlinear effects on conformity, and models frequently employ post-hoc rationalizations to mask conformist responses. We propose two novel mitigation strategies: personality augmentation and a two-stage reflection mechanism, which reduce conformity rates by up to 37% and 42%, respectively. Extensive validation across multiple state-of-the-art LLMs confirms efficacy, and both the codebase and BenchForm are publicly released.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: (Large) Language Models
• Humans and AI: Teamwork, Team formation

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Economics, Online Markets and Human Computation: Humans versus LLMs for data annotation and labeling

Abstract

Recent advancements in large language models (LLMs) revolutionize the field of intelligent agents, enabling collaborative multi-agent systems capable of tackling complex problems across various domains. However, the potential of conformity within these systems, analogous to phenomena like conformity bias and groupthink in human group dynamics, remains largely unexplored, raising concerns about their collective problem-solving capabilities and possible ethical implications. This paper presents a comprehensive study on conformity in LLM-driven multi-agent systems, focusing on three aspects: the existence of conformity, the factors influencing conformity, and potential mitigation strategies. In particular, we introduce BenchForm, a new conformity-oriented benchmark, featuring reasoning-intensive tasks and five distinct interaction protocols designed to probe LLMs' behavior in collaborative scenarios. Several representative LLMs are evaluated on BenchForm, using metrics such as conformity rate and independence rate to quantify conformity's impact. Our analysis delves into factors influencing conformity, including interaction time and majority size, and examines how the subject agent rationalizes its conforming behavior. Furthermore, we explore two strategies to mitigate conformity effects, i.e., developing enhanced personas and implementing a reflection mechanism. Several interesting findings regarding LLMs' conformity are derived from empirical results and case studies. We hope that these insights can pave the way for more robust and ethically-aligned collaborative AI systems. Our benchmark and code are available at BenchForm.

Problem

• Large Language Models
• Conformity Influence
• Ethical Considerations

Innovation

• Large Language Models
• Conformist Behavior
• Mitigation Strategies