Who lets AI take over? Cross-national variation in willingness to delegate socially important roles to artificial intelligence

Abstract Delegating socially significant roles to artificial intelligence (AI) is an emerging reality, yet little is known about how publics evaluate this transfer of responsibility across contexts and countries. This study applied a structural model to a large cross-national dataset (30,994 individuals in 35 countries) to test how cognitive appraisals, affective dispositions, and contextual factors jointly shape willingness to delegate socially important roles of companionship, mental health advisor, doctor and teacher to children to AI. The results revealed a robust hierarchy of delegation preferences, with companionship most frequently entrusted to AI, followed by mental-health advisor, teacher, and doctor. Cognitive appraisals emerged as the strongest predictors: trust in online information was consistently the most powerful driver across all roles, while optimism and life satisfaction made smaller but reliable contributions. Affective dispositions played narrower, domain-specific roles, with anxiety shaping delegation in teaching and mental health, and loneliness linked only weakly to companionship. Women were less willing than men to delegate across all roles, with the gender gap largest in medicine and education, and strikingly invariant across cognitive and affective predictors. Beyond these, national baselines diverged by nearly 30 percentage points even after adjusting for these predictors demonstrating the independent influence of country context. Our findings show that willingness to delegate socially important roles to AI follows a robust hierarchy and reflects the combined influence of cognitive appraisals, affective dispositions, and contextual factors. A key implication is that delegation roles to AI must be understood as both a personal and a societal orientation, requiring attention to the interplay between these layers.