One Model, Many Morals: Uncovering Cross-Linguistic Misalignments in Computational Moral Reasoning

2025-09-25

AI Summary

Large language models (LLMs), pretrained predominantly on English data, exhibit systematic biases and inconsistencies in moral judgment across multilingual and multicultural contexts. Method: We introduce the first cross-cultural moral reasoning benchmark covering five languages, employing zero-shot evaluation to uncover language-specific judgment disparities; we propose a structured taxonomy of moral reasoning errors and integrate pretraining data provenance analysis to identify cultural misalignment as the primary source of bias. Contribution/Results: This work provides the first empirical evidence that language is not merely an input modality but a constitutive factor shaping moral cognition. Our findings establish a methodological foundation and empirical basis for culturally aware AI design, advancing moral alignment research from monolingual toward genuinely multicultural paradigms. The benchmark, error taxonomy, and causal analysis collectively enable rigorous, cross-linguistic evaluation of ethical reasoning in LLMs.

Technology Category

• Natural Language Processing: Ethics — Bias, Fairness, Transparency & Privacy
• Philosophy and Ethics of AI: Bias, Fairness & Equity
• Machine Learning: Ethics, Bias, and Fairness

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
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions

Abstract

Large Language Models (LLMs) are increasingly deployed in multilingual and multicultural environments where moral reasoning is essential for generating ethically appropriate responses. Yet, the dominant pretraining of LLMs on English-language data raises critical concerns about their ability to generalize judgments across diverse linguistic and cultural contexts. In this work, we systematically investigate how language mediates moral decision-making in LLMs. We translate two established moral reasoning benchmarks into five culturally and typologically diverse languages, enabling multilingual zero-shot evaluation. Our analysis reveals significant inconsistencies in LLMs' moral judgments across languages, often reflecting cultural misalignment. Through a combination of carefully constructed research questions, we uncover the underlying drivers of these disparities, ranging from disagreements to reasoning strategies employed by LLMs. Finally, through a case study, we link the role of pretraining data in shaping an LLM's moral compass. Through this work, we distill our insights into a structured typology of moral reasoning errors that calls for more culturally-aware AI.

Problem

• Uncovering moral judgment inconsistencies across languages in LLMs
• Investigating cultural misalignment in multilingual moral reasoning
• Analyzing pretraining data's role in shaping AI moral compass

Innovation

• Translated moral benchmarks into diverse languages
• Uncovered moral judgment inconsistencies across languages
• Linked pretraining data to moral reasoning errors