PATIENT TRUST IN AI-ASSISTED DIAGNOSIS AND TRIAGE: A STRUCTURED NARRATIVE REVIEW OF ETHICAL, SOCIAL, AND IMPLEMENTATION CONDITIONS

Abstract

Artificial intelligence is increasingly used in medical imaging, clinical decision support, and digital triage. However, technical performance alone does not determine whether such tools will be accepted in practice. Their implementation also depends on whether patients perceive them as understandable, safe, fair, and compatible with person-centred care. This review examines the main factors that shape patient trust in AI-assisted diagnosis and triage and discusses the social and ethical conditions that support responsible adoption. A structured narrative review design was used to synthesize recent literature on patient attitudes toward clinical AI, with particular attention to radiology, diagnostic support, communication, explainability, accountability, and emerging triage applications. The literature shows a consistent pattern of conditional acceptance. Patients are often willing to accept AI when it is used as a support tool under clinician supervision, but they are much more hesitant when AI is framed as replacing doctors in final decision-making. Trust is influenced not only by explainability, but also by validation, data quality, fairness, privacy, communication, and visible human accountability. Patients repeatedly express the wish to be informed when AI is involved in their care and to retain access to a clinician who can interpret results, answer questions, and assume responsibility. The review argues that trust in medical AI should be understood as calibrated confidence rather than simple approval or rejection. From a social-science perspective, AI in diagnosis and triage is a sociotechnical arrangement whose legitimacy depends on governance, workflow design, and the continued protection of the doctor-patient relationship.