Do Political Opinions Transfer Between Western Languages? An Analysis of Unaligned and Aligned Multilingual LLMs

2025-08-07

AI Summary

This study investigates whether political stances exhibit cross-lingual transferability in multilingual large language models (MLLMs), particularly across Western languages. Method: We prompt both unaligned and English-aligned MLLMs in five Western languages with political statements, quantifying model agreement and analyzing cross-lingual consistency. Political alignment is performed via Direct Preference Optimization (DPO) using politically annotated utterances drawn from voting-advice scenarios, across multiple model scales. Contribution/Results: Unaligned models already exhibit high cross-lingual consistency in political stance. After English alignment, all languages undergo systematic, directionally coherent, and magnitude-similar political shifts—demonstrating strong cross-lingual transferability of political orientation in MLLMs. These findings reveal that generic political alignment risks cultural coarse-graining, underscoring both the necessity and the substantial challenge of fine-grained sociolinguistic and cultural alignment.

Technology Category

• Natural Language Processing: Machine Translation, Multilinguality, Cross-Lingual NLP
• Machine Learning: Large Multimodal Models (LMMs)
• Humans and AI: Learning Human Values and Preferences

Application Category

• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Web Mining and Content Analysis: Mining multimedia, multimodal, multilingual, cross-lingual Web data

Abstract

Public opinion surveys show cross-cultural differences in political opinions between socio-cultural contexts. However, there is no clear evidence whether these differences translate to cross-lingual differences in multilingual large language models (MLLMs). We analyze whether opinions transfer between languages or whether there are separate opinions for each language in MLLMs of various sizes across five Western languages. We evaluate MLLMs' opinions by prompting them to report their (dis)agreement with political statements from voting advice applications. To better understand the interaction between languages in the models, we evaluate them both before and after aligning them with more left or right views using direct preference optimization and English alignment data only. Our findings reveal that unaligned models show only very few significant cross-lingual differences in the political opinions they reflect. The political alignment shifts opinions almost uniformly across all five languages. We conclude that in Western language contexts, political opinions transfer between languages, demonstrating the challenges in achieving explicit socio-linguistic, cultural, and political alignment of MLLMs.

Problem

• Examines political opinion transfer across Western languages in MLLMs
• Assesses cross-lingual differences before and after political alignment
• Explores challenges in socio-linguistic and political alignment of MLLMs

Innovation

• Analyzing political opinion transfer in multilingual LLMs
• Using direct preference optimization for alignment
• Evaluating models before and after alignment shifts