Epistemic authority and medical AI: epistemological differences and challenges in medical practice

Recent and ongoing advances in medical AI promise to revolutionise medicine by improving the accuracy, speed, and efficiency of clinical care. These promises are responses to the continuous quest of modern medicine to eliminate uncertainty and find answers to crucial questions of diagnosis, prognosis and treatment, while the impressive reported results of medical AI have raised the question of whether medical AI can be perceived as an epistemic authority that challenges the authority of doctors. In this paper, we examine this question by approaching it from the standpoint of what epistemic goods medical AI can offer, or else, what medical AI can claim to "know". Using Popowicz' account of epistemic authority in medical practice, which he locates in the scientific method that underpins the practice, we argue that medical AI uses a different scientific method to the one that has given rise and forms the epistemic foundations of traditional western medicine, and this presents a problem. As long as we are seeking not only statistically accurate correlations, but empirically grounded causations in medicine, AI cannot be treated as an epistemic authority in this field. We conclude that until medical practice finds ways to successfully incorporate such epistemological differences, medical AI should submit to the epistemic authority of medical practice and take its place on the long list of important and useful epistemic tools doctors can use to improve the health of patients.