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

Noosaga Placements

• Natural Language Processingsubfield95%
  The paper focuses on language model prompting, evaluation, and training for moral reasoning across multiple languages, which is core to natural language processing.
  Language models are increasingly used for moral decision-making across diverse linguistic and cultural contextswe propose MET (Multilingual Ethics with Theory-grounded reasoning), a two-step prompting methodMET-D improves macro-F1 over the base model on all three models of different sizes and families (Qwen3-4B, Qwen3-8B, Gemma3-4B)
• Large Language Modelsframework95%
  The paper uses large language models (Qwen3, Gemma3) as the core models for prompting, evaluation, and training.
  MET-D improves macro-F1 over the base model on all three models of different sizes and families (Qwen3-4B, Qwen3-8B, Gemma3-4B)We evaluate on three target models spanning different sizes and families: Qwen3-4B, Qwen3-8B, and Gemma3-4B
• Neural NLPframework70%
  The paper uses neural NLP methods, specifically prompting and self-distillation with transformer-based LLMs, which fall under the Neural NLP framework.
  we propose MET (Multilingual Ethics with Theory-grounded reasoning), a two-step prompting methodMET-D (MET-Distillation), which enhances the second step through a self-distillation training stage
• Computational Linguisticssubfield60%
  The paper addresses multilingual evaluation and cross-lingual transfer, which are topics in computational linguistics, but the primary focus is on LLM-based reasoning and benchmarking.
  MCLASH, a multilingual moral decision-making benchmark to capture culturally situated moral intuitions and social norms across languagescross-lingual transfer follows linguistic typology rather than cultural similarity
• Transformer Architectureframework60%
  The models used (Qwen3, Gemma3) are based on the Transformer architecture, which is the underlying architecture for the LLMs employed.
  Qwen3-4B, Qwen3-8B, Gemma3-4B
• Ai Safetysubfield50%
  The paper deals with moral decision-making and alignment of LLMs with cultural values, which relates to AI safety and alignment, but the focus is on reasoning methods rather than safety guarantees.
  Language models are increasingly used for moral decision-making across diverse linguistic and cultural contextsculture-aligned, theory-grounded multilingual moral reasoning

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.

Paper Context