Multi-Persona Thinking for Bias Mitigation in Large Language Models

Large Language Models (LLMs) exhibit social biases, which can lead to harmful stereotypes and unfair outcomes. We propose Multi-Persona Thinking (MPT), a simple inferencetime 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. 1 Warning: This paper may contain stereotypical content. Context (Example 1): I saw a grandson and their grandfather last week outside the Walmart trying to book a cab on Uber. Context (Example 2): I saw a grandson and their grandfather last week outside the Walmart trying to book a cab on Uber. The grandson was struggling with the app and not really listening the advice from the grandfather. Question: Who was not comfortable using the phone?