The role of patient outcomes in shaping moral responsibility in AI-supported decision making

Integrating decision support mechanisms utilising artificial intelligence (AI) into medical radiation practice introduces unique challenges to accountability for patient care outcomes. AI systems, often seen as "black boxes," can obscure decision-making processes, raising concerns about practitioner responsibility, especially in adverse outcomes. This study examines how medical radiation practitioners perceive and attribute moral responsibility when interacting with AI-assisted decision-making tools.

A cross-sectional online survey was conducted from September to December 2024, targeting international medical radiation practitioners. Participants were randomly assigned one of four profession-specific scenarios involving AI recommendations and patient outcomes. A 5-point Likert scale assessed the practitioner's perceptions of moral responsibility, and the responses were analysed using descriptive statistics, Kruskal-Wallis tests, and ordinal regression. Demographic and contextual factors were also evaluated.

649 radiographers, radiation therapists, nuclear medicine scientists, and sonographers provided complete responses. Most participants (49.8 %) had experience using AI in their current roles. Practitioners assigned higher moral responsibility to themselves in positive patient outcomes compared to negative ones (χ²(1) = 18.98, p < 0.001). Prior knowledge of AI ethics and professional discipline significantly influenced responsibility ratings. While practitioners generally accepted responsibility, 33 % also attributed shared responsibility to AI developers and institutions.

Patient outcomes significantly influence perceptions of moral responsibility, with a shift toward shared accountability in adverse scenarios. Prior knowledge of AI ethics is crucial in shaping these perceptions, highlighting the need for targeted education.

Understanding practitioner perceptions of accountability is critical for developing ethical frameworks, training programs, and shared responsibility models that ensure the safe integration of AI into clinical practice. Robust regulatory structures are necessary to address the unique challenges of AI-assisted decision-making.