Trust in artificial moral advisors across cultures

Authors:

Abstract
Developers of artificial intelligence are already building prototypes for artificial moral advisors: autonomous systems designed to provide humans with recommendations on ethical issues. Yet it remains unclear whether people will trust and adopt these technologies. In a large-scale cross-cultural experiment (12 countries; N = 6,896), we investigate perceptions of human and artificial moral advisors who give advice in moral dilemmas based on competing ethical principles. We show that people trust artificial advisors less than human ones, especially when their advice aligns with utilitarian rather than deontological principles. Nonetheless, people update their own moral judgments in the direction of advice from AI advisors, and do so more when the advice is deontological. Our findings raise challenges for the design, development, and deployment of artificial moral advisors and pose serious questions about the potential influence of machines on human morality, with implications for AI governance.