PHAnToM: Persona-based Prompting Has An Effect on Theory-of-Mind Reasoning in Large Language Models

TL;DR

This paper investigates how persona-based prompting affects large language models' ability to perform social-cognitive reasoning, revealing that role-playing prompts can influence Theory-of-Mind reasoning accuracy and highlighting potential risks in prompt engineering.

Contribution

It provides empirical evidence linking role-playing prompts to variations in ToM reasoning in LLMs, grounded in psychological theory, and warns about possible errors from persona adoption.

Findings

Role-playing prompts influence ToM reasoning performance.

Models adopting specific personas may make social-cognitive reasoning errors.

Performance differences are linked to socially-motivated prompting variations.

Abstract

The use of LLMs in natural language reasoning has shown mixed results, sometimes rivaling or even surpassing human performance in simpler classification tasks while struggling with social-cognitive reasoning, a domain where humans naturally excel. These differences have been attributed to many factors, such as variations in prompting and the specific LLMs used. However, no reasons appear conclusive, and no clear mechanisms have been established in prior work. In this study, we empirically evaluate how role-playing prompting influences Theory-of-Mind (ToM) reasoning capabilities. Grounding our rsearch in psychological theory, we propose the mechanism that, beyond the inherent variance in the complexity of reasoning tasks, performance differences arise because of socially-motivated prompting differences. In an era where prompt engineering with role-play is a typical approach to adapt LLMs to new contexts, our research advocates caution as models that adopt specific personas might potentially result in errors in social-cognitive reasoning.