Influencing Officials' Adoption of AI for Risk Decision-Making: An Experimental Study in Emergency Contexts

As artificial intelligence (AI) becomes more integrated into public decision-making, its anthropomorphic features raise new questions about trust, accountability, and risk in high-stakes emergency contexts. Existing research highlights the importance of cognitive alignment between algorithmic outputs and bureaucratic reasoning, yet little is known about how AI's human-like cognitive and emotional cues shape officials' behavioral adoption. Drawing on AI anthropomorphism and dual-process theory, this study proposes a dual-path trust model linking cognitive congruence and emotional empathy in AI recommendations to officials' adoption decisions. Using a 2×2 between-subjects experiment with 322 Chinese emergency officials, the findings show that cognitive congruence has a strong positive effect on adoption, while emotional empathy has a weaker but independent effect. These results reveal a structural paradox: while emotional empathy can increase initial acceptance, only cognitive congruence reliably enhances adoption by providing the defensible rationale needed to mitigate perceived liability risks and operational uncertainty in high-stakes crises. The study offers implications for designing transparent, accountable, and trustworthy AI systems that support defensible decision-making in emergency management.