Ruleset Integrity

A sealed record proves what a check was judged against. This proves whether anyone has looked at it since.

Every check we run is tied to an exact, hashed version of the rules in force at that moment, so nobody can quietly change the rules and claim they always said this. That proves consistency. It does not prove the rules were ever right, or that anyone has checked lately. This page is the answer to that second question: who reviewed the current ruleset, and when.

Loading…

This is a narrow claim, deliberately. It does not certify the rules are correct, only that a named person has looked at this exact version and can be asked why it's still current. See boundary authorization records for the same idea applied to individual decisions rather than the rule set itself.

Firing rate is not the same evidence as miss rate.

A review above answers whether someone looked at the rules again. It doesn't answer whether the rules were ever the right rules. That can't be tested by writing a condition in the rules' own vocabulary, since a category list can only recognise the cases it already has a category for. The only real test is a back-test: take incidents from a source we don't control, regulator rulings, not our own results, and check how many the current category list would actually have caught.

Loading…