ABSTRACT
While AI is reshaping public services, the acceptance of diverse AI-based public service innovations by citizens has not been fully examined. This study conducted a video-based experiment involving 2254 citizens and four types of AI: mechanical, empathetic, analytical, and intuitive. The results show that mechanical AI was significantly more accepted by citizens, whereas empathetic AI was significantly less accepted. Acceptance of the other two types of AI was moderate, with no significant differences found between them. AI types affected acceptance through two paths: perceived usefulness and novelty. The mediating effect of AI’s perceived usefulness was found to be stronger than its novelty. Furthermore, citizens’ trust in the government positively moderated the impact of the perceived novelty of AI. Thus, we recommend that governments adopt a gradual promotion strategy when introducing AI-based public services.