Lessons Without Borders? Evaluating Cultural Alignment of LLMs Using Multilingual Story Moral Generation

2026-04-09

AI Summary

This work addresses the limited sensitivity of large language models to cultural variations in human moral judgment during cross-cultural narrative understanding. It proposes “multilingual story-based moral generation” as a novel evaluation paradigm for dynamic cultural alignment, leveraging human-authored moral judgments across 14 language–culture pairs. Employing semantic similarity metrics, human preference surveys, and a values classification framework, the study systematically evaluates state-of-the-art models including GPT-4o and Gemini. Results reveal that while model-generated moral statements exhibit semantic proximity to human responses and are often preferred by annotators, they display significantly lower cross-linguistic variability than human judgments, tending instead to converge on universalist values and underrepresent cultural diversity. This research moves beyond static benchmarks by introducing narrative moral generation as a first-of-its-kind approach to assessing cultural alignment in AI systems.

Technology Category

• Humans and AI: Learning Human Values and Preferences
• Philosophy and Ethics of AI: Morality & Value-based AI
• Natural Language Processing: (Large) Language Models

Application Category

• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• Social Networks and Social Media: Generative AI / large language models and their impact on social systems
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions

Abstract

Stories are key to transmitting values across cultures, but their interpretation varies across linguistic and cultural contexts. Thus, we introduce multilingual story moral generation as a novel culturally grounded evaluation task. Using a new dataset of human-written story morals collected across 14 language-culture pairs, we compare model outputs with human interpretations via semantic similarity, a human preference survey, and value categorization. We show that frontier models such as GPT-4o and Gemini generate story morals that are semantically similar to human responses and preferred by human evaluators. However, their outputs exhibit markedly less cross-linguistic variation and concentrate on a narrower set of widely shared values. These findings suggest that while contemporary models can approximate central tendencies of human moral interpretation, they struggle to reproduce the diversity that characterizes human narrative understanding. By framing narrative interpretation as an evaluative task, this work introduces a new approach to studying cultural alignment in language models beyond static benchmarks or knowledge-based tests.

Problem

• cultural alignment
• multilingual story moral generation
• cross-linguistic variation
• value diversity
• narrative interpretation

Innovation

• multilingual story moral generation
• cultural alignment
• large language models
• cross-linguistic variation
• narrative interpretation