Whose values are artificial intelligence models aligning with? How culture shapes people's normative expectations of artificial intelligence value

Abstract

With the rapid development of artificial intelligence (AI), growing attention has been paid to the role of culture in shaping AI values, yet existing research has rarely provided a systematic synthesis of both human universals and cultural differences in people's normative expectations of AI. Our study reveals both human universals and cultural differences among AI values. The findings indicate widespread cross-cultural commonality in the pursuit of values such as safety and universalism, as well as shared ethical standards concerning privacy, transparency, fairness, justice, and accountability. Moreover, cultural differences are evident in attitudes, behaviors, and policy orientations toward the application and regulation of AI across cultural contexts. In addition, we discuss the vital role of implicit cultural beliefs and cultural norms in the ethical supervision and practical applications of AI systems in human society. Future work should further explore developing and iterating algorithms for diverse culturally informed application scenarios, thereby both promoting the globalization of AI systems and meeting diverse cultural psychological demands to ultimately improve the well-being of individuals and groups and humanity as a whole.