Multi-Persona Thinking for Bias Mitigation in Large Language Models

TL;DR

This paper introduces Multi-Persona Thinking (MPT), a simple inference-time framework that mitigates social biases in large language models by encouraging reasoning from multiple contrasting perspectives.

Contribution

The paper proposes MPT, a novel inference-time method that reduces bias by integrating multiple social viewpoints during reasoning in large language models.

Findings

MPT achieves lower bias scores than existing prompting methods.

MPT maintains the core reasoning ability of the models.

Evaluations on two bias benchmarks demonstrate effectiveness.

Abstract

Large Language Models (LLMs) exhibit social biases, which can lead to harmful stereotypes and unfair outcomes. We propose \textbf{Multi-Persona Thinking (MPT)}, a simple inference-time framework that reduces social bias by encouraging reasoning from multiple perspectives. MPT guides the model to consider contrasting social identities, such as male and female, together with a neutral viewpoint. These viewpoints then interact through an iterative reasoning process to identify and correct biased judgments. This design transforms the potential weakness of persona assignment into a mechanism to mitigate bias. We evaluate MPT on two widely used bias benchmarks with both open-source and closed-source models. Our results show that MPT achieves a lower bias than the existing prompting-based methods while maintaining the core reasoning ability.