Processing and Witnessing in Algorithmic Radiology

This article examines the ethical implications of algorithmic processing in radiology through the distinction between processing and witnessing, two dimensions of medical practice long fused within the single figure of the physician. It argues that as algorithmic systems increasingly absorb the processing component of radiological work (pattern recognition, quantification, classification), witnessing does not disappear but relocates into three mediated channels: the language of the written report, the dialogue between radiologist and referring clinician, and the act of institutional endorsement that follows algorithmic output. Drawing on Levinasian ethics of address and the concept of homo narratus, the article foregrounds what algorithmic processing cannot inherit: the condition of being claimed by another person's vulnerability, and of answering that claim through language, dialogue, and signature rather than through output alone. Radiology, treated here as a limit case in which processing has advanced furthest, renders this distinction unusually visible before it becomes visible elsewhere in clinical practice. The article addresses a range of objections, including the claim that narrative competence is itself reducible to computational processing, and considers the implications of this argument for the future of medical education, without proposing a specific curricular model. It concludes that as processing becomes increasingly automated across medical specialties, the distinctly human contribution of medicine is likely to rest ever more heavily on witnessing.