Piotr Wiliński ¹, Robert Ranisch¹

¹ Juniorprofessur für Medizinische Ethik mit Schwerpunkt auf Digitalisierung, Fakultät für Gesundheitswissenschaften Brandenburg, Universität Potsdam

The implementation of AI in healthcare, e.g., in diagnostic and imaging processes, medical decision making, robots in care and surgery, etc., carries both great promise, as well as serious risks. The increasing autonomy of these technologies challenges existing accountability practices within the patient-physician relationship. Where medical personnel cannot be justly and fairly held responsible for outcomes caused by autonomous technologies, “responsibility gaps” may emerge. It has been argued that the moral permissibility of AI technologies in healthcare relies upon our ability to successfully close these gaps. To that end, some authors propose responsibilisation: an approach whereby physicians can voluntarily assume responsibility for outcomes caused by AI qua their professional role. The contribution of the present paper is a moral substructure which allows responsibilisation practices to be meaningful, i.e., based upon reasonable and morally salient conditions. We argue that although physicians might not be morally responsible for outcomes caused by AI in a traditional sense, they can – and perhaps should - nonetheless assume responsibility by expanding their existing moral obligations towards the people in their care (and justifiably so). These existing moral obligations stem from their professional role as caregivers, and are directly built into the patient-physician relationship. We propose two conditions for meaningful responsibilisation: (1) expertise and (2) professional role. The first condition concerns the epistemic standing empowering physicians to fulfil their duties; the second condition regards the strict legal and moral standards concerning the various duties (and their limits) imposed on physicians depending on their specialisation, as well as ties directly into the sort of person physicians are expected to be – a condition upon which the nature of the patient-physician relationship is constituted. Considering the conditions for meaningful and reasonable responsibilisation in medical practice is important given the considerable potential to scapegoat medical personnel for AI-caused outcomes or to ‘randomise’ the process of responsibility ascription.