LLMs Show Judgment-Consequence Gap in Medical Resource Allocation
August 7, 2026
LLMs recognize patient responsibility for health-harming behaviors but fail to apply this judgment to resource allocation, defaulting to random distribution instead of human-like selective allocation. This gap persists across various model families when presented with clinical vignettes involving scarce medical resources.
HOW THIS AFFECTS YOU
●
policyThis suggests LLMs may require specific alignment to ensure moral reasoning translates into consistent policy execution.
●
healthYou must account for unexpected resource allocation patterns when deploying LLMs in clinical decision support.