The Ethics of Counting Risk: Moral Consistency and Reference Classes under Repeated Decisions

A lower reported risk need not mean that those already exposed are safer. An institution can improve its pooled expected harm rate by adding genuinely exposed lower-risk cases, while leaving the original cohort’s outcome distribution unchanged and increasing expected harms. Expansion may still be beneficial. We argue that its benefits must enter through an explicit justification: where the original cohort’s prospects and the independently relevant reasons for imposing its risks are unchanged, a better pooled score cannot itself answer that cohort’s objection. Lifetime protection faces a complementary difficulty: service and prospect requirements can become jointly infeasible even while a failure-free outcome remains possible. Neither a favorable realization nor a change of accounting scope resolves that conflict. We defend a plural ledger organized around the claims an institution must answer. It distinguishes improved protection, improved justification, and improved reporting; requires compatible accounts of shared facts; and preserves answerability through aggregation and justified revision. Statistical accuracy cannot substitute for a reason addressed to those who bear the risk.