Appel1 argues that artificial intelligence healthcare resource allocation (AI-HRA) will make harms to ‘invisible victims’—defined initially as those ‘[…] not aware [of being shortchanged by trade-offs]’1 but later broadened to include those shortchanged ‘[… without] full awareness of the implications’ on the part of decision makers11—‘visible’, allowing HRA to take clearer stock of salient facts. AI-HRA is also proposed as a tool to mitigate biases and misinformation surrounding HRA, adapt quickly to changing realities, and democratise participation in policy making by making relevant insights available to non-experts. In these applications, the purported objective accuracy of AI-HRA is proposed as a source of a sort of practical and ethical challenge for which our current systems of evaluation are not prepared—the need to handle explicit trolley problems at all stages of HRA.
While AI tools are a growing force at all levels of HRA, and…