Cross-cultural value alignment frameworks for responsible AI governance: Evidence from China-West comparative analysis

TL;DR

This paper develops a comprehensive framework for evaluating and comparing cross-cultural value alignment in large language models from China and the West, revealing key challenges and differences in cultural fidelity and model performance.

Contribution

It introduces a multi-method auditing platform and provides a comparative analysis of over 20 models, highlighting the impact of architecture and training data on cultural alignment.

Findings

Universal challenges in value stability and demographic representation.

China models focus on multilingual data for cultural context.

Western models show architectural experimentation but U.S.-centric biases.

Abstract

As Large Language Models (LLMs) increasingly influence high-stakes decision-making across global contexts, ensuring their alignment with diverse cultural values has become a critical governance challenge. This study presents a Multi-Layered Auditing Platform for Responsible AI that systematically evaluates cross-cultural value alignment in China-origin and Western-origin LLMs through four integrated methodologies: Ethical Dilemma Corpus for assessing temporal stability, Diversity-Enhanced Framework (DEF) for quantifying cultural fidelity, First-Token Probability Alignment for distributional accuracy, and Multi-stAge Reasoning frameworK (MARK) for interpretable decision-making. Our comparative analysis of 20+ leading models, such as Qwen, GPT-4o, Claude, LLaMA, and DeepSeek, reveals universal challenges-fundamental instability in value systems, systematic under-representation of younger demographics, and non-linear relationships between model scale and alignment quality-alongside divergent regional development trajectories. While China-origin models increasingly emphasize multilingual data integration for context-specific optimization, Western models demonstrate greater architectural experimentation but persistent U.S.-centric biases. Neither paradigm achieves robust cross-cultural generalization. We establish that Mistral-series architectures significantly outperform LLaMA3-series in cross-cultural alignment, and that Full-Parameter Fine-Tuning on diverse datasets surpasses Reinforcement Learning from Human Feedback in preserving cultural variation...