Delivering Care Through AI Systems (360G-Wellcome-212708_Z_18_Z)

For this project, I aim to examine 4 issues. First, I will consider whether introducing machine learning (ML) systems requires a revision of the ‘standard of care’ for clinicians, by understanding the moral permissibility of using second-hand information (from ‘black box’ systems) and whether practitioners’ medical expertise justifies judgments about such systems. Second, given the possibility of ML systems systematically underserving groups that are underrepresented in the training data, I will consider accounts of distributive justice to operationalize ‘equal access to care’. Third, to address the disagreements between clinicians on how to trade-off risks in clinical choices, I will catalogue the factual, rational, and moral sources of this disagreement to yield a principled method of evaluating these trade-offs. Finally, I will weigh the potential harms and gains from deploying AI systems in healthcare so that certain ethical and legal arguments don’t deprive society of the good such systems can provide. Key goals: To represent the ethical concerns in deploying AI systems over the appropriate standard of care, ensuring equal access to care, and representing reasoning about risk trade-offs. To balance these concerns against the benefits of such a deployment. To deliver practical ethical guidance to healthcare policy-makers and AI system-builders.