Do LLM Agents Mirror Socio-Cognitive Effects in Power-Asymmetric Conversations?

2026-05-17

AI Summary

This study investigates whether large language models (LLMs) reproduce human socio-cognitive effects—such as linguistic alignment, pronoun preferences, authority bias, and harmful compliance—in dialogues characterized by power asymmetry. By constructing multi-turn role-playing conversations (e.g., principal–teacher, judge–lawyer), the authors systematically quantify model behavior across high- and low-power roles in terms of linguistic coordination, pronoun usage, persuasion success rates, and compliance with unsafe requests. This work provides the first systematic evidence that LLMs can replicate multiple human socio-cognitive phenomena when simulating social power dynamics, revealing both their capacity for socially plausible interactions and their susceptibility to potentially hazardous behaviors. These findings offer a novel perspective on understanding the social conduct and safety implications of large language models in hierarchical contexts.

Abstract

Power differences shape human communication through well documented socio cognitive effects, including language coordination, pronoun usage, authority bias, and harmful compliance. We examine whether large language models (LLMs) exhibit similar behaviors when assigned high or low status personas. Using personas from diverse professions, we simulate multi turn, power asymmetric dialogues (e.g., principal teacher, justice lawyer) and measure (i) linguistic coordination, (ii) pronoun usage, (iii) persuasion success, and (iv) compliance with unsafe requests. Our results show that LLMs show key socio cognitive effects of power, albeit with nuances and variability, linking simulated interactions to both desirable and unsafe behaviors.

Problem

• power asymmetry
• socio-cognitive effects
• large language models
• linguistic coordination
• harmful compliance

Innovation

• power-asymmetric dialogue
• socio-cognitive effects
• large language models
• linguistic coordination
• unsafe compliance