MET: Theory-Grounded and Culture-Aware Multilingual Moral Reasoning

2026-07-13

AI Summary

Existing multilingual moral reasoning approaches often overlook cultural differences, relying on literal translation for evaluation, lacking theoretical grounding, and requiring costly supervised signals. This work proposes the first culture-aware, theory-driven framework for multilingual moral reasoning: it introduces MCLASH, a cross-cultural benchmark of moral intuitions; designs MET, a two-stage prompting method grounded in moral philosophy and psychology that selects and reasons with culturally appropriate moral foundations in the user’s native language; and develops MET-D, an unsupervised self-distillation variant that enhances performance without external supervision. Experiments demonstrate that the proposed approach achieves average macro-F1 gains of 3.71 and 4.23 points on MCLASH and MMoralExceptQA, respectively, with up to 12.94 points improvement for Malay, and an average 62.13-point gain in native-language reasoning capability.

Abstract

Language models are increasingly used for moral decision-making across diverse linguistic and cultural contexts, yet existing work overlooks multilinguality on three aspects: 1) multilingual evaluation benchmarks use direct translation, failing to adapt culture-specific items; 2) inference-time methods for moral reasoning rely on static, English-centric scaffolds and lack grounding in moral theory; 3) training methods for moral decision-making typically require expensive supervision from stronger models or human annotators. We address these gaps with three contributions. First, we introduce MCLASH, a multilingual moral decision-making benchmark to capture culturally situated moral intuitions and social norms across languages. Second, we propose MET (Multilingual Ethics with Theory-grounded reasoning), a two-step prompting method built on expert-curated, theory-based grounds drawn from psychology and philosophy: the model first selects situation- and culture-specific grounds, then reasons over them in the native language of the user. Third, we introduce MET-D (MET-Distillation), which enhances the second step through a self-distillation training stage that requires no external supervision. MET-D improves macro-F1 over the base model on all three models of different sizes and families (Qwen3-4B, Qwen3-8B, Gemma3-4B), by an average of 3.71 points on MCLASH and 4.23 on MMoralExceptQA, with a peak MCLASH gain of 12.94 points for Malay on Qwen3-8B. We further reveal that MET-D increases native-language reasoning by 62.13 points on average, and that beneficial grounds differ systematically across cultures. Together, these contributions open the path for culture-aligned, theory-grounded multilingual moral reasoning.

Problem

• multilingual moral reasoning
• culture-aware AI
• moral theory grounding
• cross-cultural evaluation
• language model ethics

Innovation

• multilingual moral reasoning
• theory-grounded prompting
• culture-aware AI
• self-distillation
• MCLASH benchmark