Artificial Intelligence Perceptions and Technostress in Staff Radiologists: The Mediating Role of Artificial Intelligence Acceptance and the Moderating Role of Self-Efficacy

Abstract

This study examined how perceptions of artificial intelligence (AI) relate to technostress in healthcare professionals, testing whether AI acceptance mediates this relationship and whether self-efficacy moderates the formation of acceptance. Seventy-one participants completed measures of Perceptions of AI (Shinners), AI Acceptance (UTAUT), Self-Efficacy, and four technostress outcomes: Technostress Overall, Techno-Overload, Techno-Complexity/Insecurity, and Techno-Uncertainty. Conditional process analyses (PROCESS Model 7; 5000 bootstrap samples) were performed controlling for sex, age (years), and professional role (radiology residents, attending radiologists, PhD researchers). Perceptions of AI were directly and positively associated with Technostress Overall (b = 0.57, p = 0.003), Techno-Overload (b = 0.58, p = 0.008), and Techno-Complexity/Insecurity (b = 0.83, p < 0.001), but not with Techno-Uncertainty (b = −0.02, p = 0.930). AI Acceptance negatively predicted the same three outcomes (e.g., Technostress Overall b = −0.55, p = 0.004), and conditional indirect effects indicated significant negative mediation at low, mean, and high self-efficacy for these three outcomes. Self-efficacy moderated the Perceptions → Acceptance path (interaction b = −0.165, p = 0.028), with a stronger X→M effect at lower self-efficacy, but indices of moderated mediation were not significant for any outcome. The results suggest that perceptions of AI exert both demand-like direct effects and buffering indirect effects via acceptance; implementation should therefore foster acceptance, build competence, and address workload and organizational clarity.