Introducing the ethical-epistemic matrix: a principle-based tool for evaluating artificial intelligence in medicine

Abstract While there has been much discussion of the ethical assessment of artificial intelligence (AI) in medicine, such work has rarely been combined with the parallel body of scholarship analyzing epistemic implications of AI. This paper proposes a method for joint evaluation of AI’s ethical and epistemic implications in medicine that draws on the principle-oriented tradition in bioethics and the consequent ‘ethical matrix’ approach to assessing novel technologies. It first introduces principle-based approaches as specific tools for ethical assessment of AI in medicine and other domains that are contrasted with the lack of comparable epistemic principles that would govern AI evaluation in medicine. In the next section, the ethical matrix is explained as a well-established principle-based tool in applied ethics that has had some limited applications to near-term implications of AI in medicine and elsewhere that can be strengthened, I suggest, using epistemic principles. To this end, the following section looks to the philosophy of science for relevant epistemic principles, identifying ‘accuracy’, ‘consistency’, ‘relevance’, and ‘instrumental efficacy’ as a provisional set for technology evaluation. The next section articulates the relevance of these epistemic principles to AI in medicine by highlighting conventional standards that have already been applied in AI, epistemology, and the medical sciences. Before concluding, the paper then defines and defends the possibility of an ‘ethical-epistemic matrix’ for the application of these epistemic principles alongside established ethical principles to a selection of stakeholder groups: patients, clinicians, developers, and the public.