How Citizen Experiences Shape Attitudes Toward AI in Public Service Delivery: A Survey Experiment on Attitudinal Shifts

Abstract As governments increasingly adopt artificial intelligence (AI) in public service delivery, understanding how citizens respond to AI-driven decisions is critical. Using survey experiment data ( N = 3,010), this study examines how citizens’ attitudes toward AI change after direct experiences with AI or human decision-makers in tax and SNAP policy contexts. The findings reveal a dual-process effect: exposure to AI increases beliefs in its performance capabilities but reduces overall perceptions relative to human interactions. The study also identifies a strong negativity bias, whereby negative AI experiences disproportionately erode trust, underscoring the need for phased AI deployment and robust expectation management.