Bias Beyond Borders: Political Ideology Evaluation and Steering in Multilingual LLMs
This addresses fairness and ideological neutrality in global AI deployment, focusing on cross-lingual consistency and mitigation, but it is incremental as it builds on existing steering methods.
The paper tackled the problem of political bias in multilingual LLMs by evaluating it across 50 countries and 33 languages and introducing a post-hoc mitigation framework called Cross-Lingual Alignment Steering (CLAS), which achieved substantial bias reduction with minimal degradation in response quality.
Large Language Models (LLMs) increasingly shape global discourse, making fairness and ideological neutrality essential for responsible AI deployment. Despite growing attention to political bias in LLMs, prior work largely focuses on high-resource, Western languages or narrow multilingual settings, leaving cross-lingual consistency and safe post-hoc mitigation underexplored. To address this gap, we present a large-scale multilingual evaluation of political bias spanning 50 countries and 33 languages. We introduce a complementary post-hoc mitigation framework, Cross-Lingual Alignment Steering (CLAS), designed to augment existing steering methods by aligning ideological representations across languages and dynamically regulating intervention strength. This method aligns latent ideological representations induced by political prompts into a shared ideological subspace, ensuring cross lingual consistency, with the adaptive mechanism prevents over correction and preserves coherence. Experiments demonstrate substantial bias reduction along both economic and social axes with minimal degradation in response quality. The proposed framework establishes a scalable and interpretable paradigm for fairness-aware multilingual LLM governance, balancing ideological neutrality with linguistic and cultural diversity.