Professional Expertise Practice Adaptation and Sense of Agency During AI Implementation in Radiology

This study examines psychological processes during narrow-task artificial intelligence (NT-AI) implementation in radiological practice, focusing on how professionals cope with tensions arising from discrepancies between intended and experienced expertise practices with NT-AI. Such tensions may hold a particular threat to professionals’ sense of agency (SoA) and are therefore particularly important to address in a process of AI implementation. Using ethnographic methods across two Danish hospitals, we observe professionals’ typical coping strategies to core experiences of tension-based challenges: disruption, re-validation, vigilance, rejection, and practice transformation. We analyze these coping responses based on a typology of four distinct ideal types of human-AI collaboration: Parallel expertise, Forwarded expertise, Augmented expertise, and Collective expertise. Our study contributes to a nuanced understanding of human-AI collaboration by demonstrating how professionals adaptively navigate between these ideal types to transform their professional practice and reconstitute their SoA.