Who reads the receipt

Most of the effort in building accountable systems goes into the producing of records, and almost none into the reading of them. This is understandable. Producing a record is a problem you can solve with engineering: capture the inputs, freeze the rules, store enough state to replay the decision later. It is bounded, it is testable, and you can declare victory the moment the artifact exists. The reading of the record is a problem you cannot solve with engineering at all, because it is not a property of the artifact. It is a property of the institution on the other end — whether anyone there has the standing to act on what the record shows, the time to look at it, and the authority to make the looking matter. The first problem is hard. The second is harder, and we keep mistaking the first for the whole job.

A Decision Receipt that no one reads is not an accountability mechanism. It is a costume. The decision still happened the way it happened; the consequence still landed where it landed; the only thing the record changed is the institution's ability to say, truthfully, that a record exists. And "a record exists" is a sentence that does an enormous amount of reputational work while doing almost no accountability work at all. It converts the absence of scrutiny into the appearance of scrutiny. It lets an organization point at a folder and call it a conscience.

So the question I want to put at the center is not did you produce a record but who reads it, and what happens when they do. Those are the questions a receipt cannot answer about itself. They are answered by the design of the institution that surrounds it.

Accountability is not the existence of a record. It is the existence of a reader who is empowered, expected, and equipped to act on it.

Three things a reader needs

Start with standing. Someone has to be entitled to read the receipt and to do something about what it says. A record handed to a party with no power to change the outcome is a record handed to a spectator. This is the failure mode that contestability is supposed to prevent, and it is precisely the one that survives most reforms intact: organizations publish the record, congratulate themselves on transparency, and route every actual objection back to the same office that made the original decision. The reader exists, the receipt exists, and the loop closes on itself. Standing means the reader sits somewhere the decision-maker cannot simply overrule — otherwise reading the receipt is just watching a verdict get reaffirmed by the people who reached it.

Then time. Reading a real record is work, and work has to be paid for in attention. An institution that produces a thousand receipts a day and staffs one reviewer has not built accountability; it has built a backlog with a clean conscience. The receipt that is technically available but practically unread is the most common form of accountability theater there is, because it is the cheapest to produce and the easiest to defend. Nobody lied. The record was right there. It simply was never going to be opened, and everyone in the system knew it. Designing for a reader means designing for the reader's scarcity — fewer records that matter more, surfaced to someone whose job is to look, rather than a flood that guarantees no one will.

Then capacity. Even a reader with standing and time can be defeated by a record built to be technically complete and humanly opaque. If understanding the receipt requires expertise the reader does not have, the receipt has been made unreadable in a way that looks like rigor. A record's contestability is bounded by the competence of the person allowed to contest it; a receipt only the system's own engineers can interpret has quietly disqualified everyone else. The point of a receipt that carries its real evidence and its frozen rules is that a competent outsider can follow the decision without taking the system's word for anything. If the reader cannot follow it, the openness was decorative.

The reader is the design problem

Once you take the reader seriously, a lot of accountability work reveals itself as half-finished. We have gotten good at the production side — provenance, replay, the discipline of writing down what was actually in front of a decision at the moment it was made. These are real and they matter. But they are inputs to accountability, not accountability itself, and treating the receipt as the finish line lets institutions discharge an obligation by manufacturing an artifact nobody is positioned to use.

The harder design question, the one that does not yield to engineering, is who you are building the receipt for. Name that person. Give them standing the decision-maker cannot revoke. Give them a caseload small enough that reading is possible and a mandate that makes reading expected rather than heroic. Give them a record they can actually understand, and the authority to make the reading bite. Do that, and the receipt becomes what it was always supposed to be — not a souvenir of a decision, but the thing a competent, empowered reader uses to overturn it when it was wrong.

Until then, ask the question the production values are designed to forestall. Not is there a record — there is almost always a record now. Ask who reads it, what they are allowed to do about it, and what it would cost them to look. If the honest answer is no one, nothing, and everything, you are not looking at accountability. You are looking at its set dressing.

— Dispatches · Summit Cognitive