Steering Towards Fairness: Mitigating Political Bias in LLMs

TL;DR

This paper investigates political biases in large language models by analyzing internal representations and proposes a method to mitigate such biases using contrastive analysis and steering vectors.

Contribution

It introduces a novel framework for probing and mitigating political bias in decoder-based LLMs through internal activation analysis and a layer-wise mitigation approach.

Findings

Decoder LLMs encode political biases across layers.

Contrastive activation analysis reveals ideological disparities.

Mitigation via steering vectors reduces bias effectively.

Abstract

Recent advancements in large language models (LLMs) have enabled their widespread use across diverse real-world applications. However, concerns remain about their tendency to encode and reproduce ideological biases along political and economic dimensions. In this paper, we employ a framework for probing and mitigating such biases in decoder-based LLMs through analysis of internal model representations. Grounded in the Political Compass Test (PCT), this method uses contrastive pairs to extract and compare hidden layer activations from models like Mistral and DeepSeek. We introduce a comprehensive activation extraction pipeline capable of layer-wise analysis across multiple ideological axes, revealing meaningful disparities linked to political framing. Our results show that decoder LLMs systematically encode representational bias across layers, which can be leveraged for effective steering vector-based mitigation. This work provides new insights into how political bias is encoded in LLMs and offers a principled approach to debiasing beyond surface-level output interventions.