Morality Meets Machine: How Evaluator and Agent Gender Shape Human-AI Moral Judgment

As gendered AI agents are increasingly integrated into social domains, how users morally evaluate their misconduct and whether such judgments vary by gender remains underexplored. This series of studies explores how the gender of evaluators, AI agents, and patients influences moral evaluations of AI across multiple domains. Male participants tended to show greater leniency toward AI misconduct, often attributing it to technical issues in ways that may align with physical or design perspectives. Female participants exhibited a domain-sensitive pattern, showing more leniency in individualizing domains (fairness, liberty, purity) while applying more human-like moral standards in binding domains (harm, betrayal, subversion), potentially reflecting an intentional stance. Their explanations also contained richer emotional and relational language. Additionally, male evaluators seemed to project gender role expectations onto opposite-gender AI agents, judging feminized AI more harshly when harming female patients. These findings offer important implications for more equitable and effective human-computer interaction design.