Observed outcome audit

Count both kinds of wrong.

Turn later-labeled gate decisions into correct error rates and compare their cost with blanket restriction or blanket access.

FALSE POSITIVE COST

FALSE NEGATIVE COST

decision: restrict or allow. outcome: harmful or benign. Costs are relative units chosen by the analyst. Sample cases do not validate the source's 99% statement.

Load the sample or paste labeled case CSV.

—TRUE POSITIVE · RESTRICT/HARMFUL

—FALSE POSITIVE · RESTRICT/BENIGN

—FALSE NEGATIVE · ALLOW/HARMFUL

—TRUE NEGATIVE · ALLOW/BENIGN

—false-positive rate

—false-negative rate

—lowest observed cost

Awaiting labeled outcomes.