Navigating Artificial Intelligence Ambivalence in Radiology: A Conceptual Review of Cognitive, Professional, and Ethical Tensions in Hybrid Human–Algorithm Decision-Making

Abstract

Artificial intelligence (AI) is increasingly embedded within radiological practice, enhancing diagnostic accuracy and efficiency in workflows. Nonetheless, radiologists' engagement with AI is characterized by persistent cognitive, professional, and ethical tensions, beyond binary framings of trust versus resistance. These tensions can be conceptualized as AI ambivalence—a stable state in which radiologists recognize the clinical utility of AI whilst simultaneously remaining uneasy about its effects on accountability, autonomy, and justice.

A narrative, conceptual approach was taken, synthesizing empirical and theoretical literature across radiology, medical AI, human factors, ethics, and health governance. The analysis focuses on radiological practice, including image interpretation, reporting workflows, triage systems, and professional accountability. Rather than evaluating individual AI tools, the review examines how AI reshapes the conditions under which radiologists exercise clinical judgement within hybrid human–algorithm decision-making.

AI ambivalence manifests across three interrelated modalities: cognitive (epistemic uncertainty and automation bias), professional (erosion of autonomy and responsibility–authority mismatch), and ethical (tensions between efficiency, opacity, and accountability). These are sustained by four structural asymmetries in hybrid AI workflows: opacity asymmetry, responsibility asymmetry, performance ambiguity, and temporal instability. Downstream consequences include inconsistent engagement with AI outputs, variable clinical reasoning, medicolegal uncertainty, and organizational friction within health care.

AI ambivalence in radiology may be argued to represent an enduring, ethically productive feature of hybrid diagnostic practice. Targeted mitigation should aim to dampen ambivalence where it produces clinical or organizational harm, while preserving reflective skepticism to safeguard professional judgement. Understanding AI ambivalence is essential to ensuring that AI adoption in radiology supplements, rather than supplants, human expertise.