DISPATCHES · Summit Cognitive

← Tools

The Admissibility Scorecard

Rate a system, honestly.

A 0–4 maturity rubric for one property: how accountable a decision system is.

This rates accountability properties — provenance, replay, standing, contestability — not whether a decision is correct, fair, or lawful. A system can score at the top of every dimension and still be wrong on the merits in every case; what the rubric measures is whether the apparatus for catching and contesting errors exists, and how mature it is. For each of the four dimensions below, read the five maturity levels and pick the one whose criteria your system actually meets — scored from a real decision you can inspect, not the diagram of what it should preserve. It surfaces gaps; it is not a certification and not legal advice. Nothing you enter leaves this page.

By Brian C. Long · Founder, Summit Cognitive

Provenance

What the decision actually drew on — the specific inputs in front of the process at the moment it ran, recorded as they stood rather than summarized after the fact.

Pick the highest level whose criteria your system fully meets.

What an auditor checks: Pull three to five real past consequential decisions. For each, can you see the specific inputs, in the state they were in at decision time? Can you confirm they were not altered since? Where is the record silent, and is that silence acknowledged?

Replay

The ability to re-run the decision against what was known at the time and watch where it lands — to reproduce the conclusion rather than read a story about it.

Pick the highest level whose criteria your system fully meets.

What an auditor checks: Take a past consequential decision and attempt to re-run it against its preserved basis. Does it land in the same place? If not, is the divergence detected and explained, or silent? Is replay something the operator has actually done before, or only describes as possible?

Standing

A clear, established answer to who may demand the record and contest what it shows — specified in advance, not improvised when a challenge arrives.

Pick the highest level whose criteria your system fully meets.

What an auditor checks: For a given consequential decision, who is entitled to demand this record? Is the affected person on that list? Is there a forum and a remedy, or only a complaint channel? Has anyone actually obtained a record this way, or is the entitlement untested?

Contestability

Whether the record is built, from the start, to be fought — reachable and usable by the party who would dispute the decision. It is what provenance, replay, and standing amount to when seen from the standpoint of the person the decision landed on.

Pick the highest level whose criteria your system fully meets.

What an auditor checks: In the position of an affected party who believes a decision was wrong — with what the system preserves and exposes, could you build a challenge that presses on the actual basis? What specifically would stop you? Has anyone done it?

— Dispatches · Summit Cognitive