The unappealable yes

When we design for accountability, we almost always picture the same scene. Someone has been told no. The loan was declined, the claim was rejected, the application was refused, and the person on the wrong end of it wants to know why. That person is angry, and motivated, and present. They will write the email, file the complaint, demand the record. So we build the machinery of contestability to face them: the right to an explanation, the right to appeal, the record that has to hold up when challenged. The denial complains, so we build for the complaint.

But the denial is only half of what a decision system does. The other half is the approval, and the approval is also a decision. It can be made on bad evidence, against the rules, by a process that would not survive review. It can be flatly wrong. The difference is not in the quality of the reasoning — a careless yes and a careless no come from the same indifference — but in who is standing there afterward to object. And the answer, for the yes, is usually no one.

Consider who is actually harmed by a mistaken approval. The borrower extended credit they could not carry, who is grateful at the moment of the yes and ruined eighteen months later. The transaction waved through that should have been stopped, whose cost lands on a counterparty who never saw it happen. The patient told they are fine, sent home with a record that says cleared, who has no reason to contest the decision because the decision told them there was nothing to contest. In each case there is a person who was injured by an approval. In each case that person has neither the standing to challenge it nor, often, even the knowledge that a decision was made.

Why the yes escapes

An approval is built to disappear. When a decision says no, it produces friction by design — a refusal is a wall the affected party runs into, and the wall is the notice. The denial announces itself. The approval, by contrast, is the absence of an obstacle. It feels not like a decision but like the natural course of things, like nothing happening. The transaction simply cleared. The account simply opened. The patient simply went home. Nobody experiences a smooth approval as a verdict, and so nobody thinks to demand the record behind it.

This is what makes the yes unappealable, and the word matters. It is not that an appeal would fail. It is that there is no appeal to bring, because the structures of contestability were never extended to the approving side. There is no form. There is no clock running. There is, frequently, no record built to be examined at all — because a record built to be contested is a cost, and an organization spends that cost only where it expects a challenge. It expects challenges to its denials. It does not expect anyone to come knocking about a yes.

A denial is a decision that complains. An approval is a decision that disappears. We built our whole apparatus of accountability for the one that complains.

And so the asymmetry compounds. The denied party gets the explanation, the appeal, the audited trail. The party harmed by an erroneous approval gets nothing — not because their injury is smaller, but because their injury arrives late, lands on someone who was not in the room, and is never connected back to the decision that caused it. The harm of a wrong yes is real, but it is diffuse and delayed, and diffuse, delayed harms do not generate the kind of immediate, located complaint that our accountability machinery is tuned to receive.

What a record owes the other side of the decision

The principle of standing — who is entitled to contest a decision, and on what record — has to be enlarged. We have treated standing as belonging to the person the decision acted against, where acting against means refusing. But the patient sent home wrongly, the counterparty exposed by a cleared transaction, the borrower set up to fail by a too-easy yes were all acted upon, just on a delay. They have an interest in the correctness of the decision every bit as real as the rejected applicant's. What they lack is not an interest. It is a record and a route.

The remedy is not to make approvals harder. It is to make them accountable on the same terms as denials. Every consequential yes should carry the same Decision Receipt a no would carry: the evidence actually consulted, the rules active at the time, enough state to replay the decision and watch where it lands. The receipt should be produced whether or not anyone is expected to ask for it — because the whole point is that the party who will need it is not in the room and cannot ask now. The record has to exist before the harm surfaces, because by the time it surfaces the conditions that produced the decision will have moved, and reconstruction after the fact is recollection, not provenance.

This is uncomfortable, because it costs something to instrument the decisions nobody is currently fighting. But the test of an accountable system is not whether it can answer the complaints it receives. It is whether it can answer the complaints it should receive and does not — because the injured party does not know they were injured, or does not know a decision was made, or has no door to knock on. A system that produces airtight accounts of its refusals and no account at all of its approvals has not solved accountability. It has solved public relations. It has made itself defensible against the people who show up and silent toward the people who can't.

Accountability that attaches only to the no is half an account. The decision to admit, to clear, to approve, to reassure is a decision in the full sense, with a party on the far side of it who may be harmed and who deserves the same standing, the same record, the same chance to say this was wrong, and here is the proof. Build the receipt for the yes. It is the half of the account we have been leaving out — and it is the half where the person harmed is least able to demand it for themselves.

— Dispatches · Summit Cognitive