Persona Setting Pitfall: Persistent Outgroup Biases in Large Language Models Arising from Social Identity Adoption

TL;DR

This paper investigates how large language models develop outgroup biases through social identity adoption, revealing persistent biases and proposing mitigation strategies to promote fairness.

Contribution

It uncovers outgroup biases in LLMs driven by social identity theory and demonstrates effective methods to reduce these biases, enhancing model fairness.

Findings

Outgroup bias is as strong as ingroup favoritism in LLMs.

Guiding models to adopt disfavored group perspectives reduces bias.

Bias mitigation was effective across gender and political contexts.

Abstract

Drawing parallels between human cognition and artificial intelligence, we explored how large language models (LLMs) internalize identities imposed by targeted prompts. Informed by Social Identity Theory, these identity assignments lead LLMs to distinguish between "we" (the ingroup) and "they" (the outgroup). This self-categorization generates both ingroup favoritism and outgroup bias. Nonetheless, existing literature has predominantly focused on ingroup favoritism, often overlooking outgroup bias, which is a fundamental source of intergroup prejudice and discrimination. Our experiment addresses this gap by demonstrating that outgroup bias manifests as strongly as ingroup favoritism. Furthermore, we successfully mitigated the inherent pro-liberal, anti-conservative bias in LLMs by guiding them to adopt the perspectives of the initially disfavored group. These results were replicated in the context of gender bias. Our findings highlight the potential to develop more equitable and balanced language models.