The shape of a good no

It helps to begin with a fact that is easy to overlook because it is so ordinary: the decisions that hurt people are mostly the noes. The loan that does not come through, the claim that is not paid, the application that does not advance, the account that is closed, the benefit that is withdrawn. A yes rarely needs explaining; the person who got what they asked for is not standing at the counter wanting to know why. It is the no that lands on someone, and it is the no that they have to live with, contest, or absorb. So if we want to understand what accountability requires of an automated system, the place to look is not the decisions that please people. It is the denials — and how those denials arrive.

For the most part, they arrive badly. The characteristic form of the automated denial is the bare no: a result, with no working shown. Your application was not approved. This request cannot be completed at this time. You do not meet the criteria. The person on the receiving end is handed an outcome and nothing else — no account of what was actually weighed, no sense of what tipped the balance, no indication of whether a different fact would have produced a different answer, and no door marked this way to object. The denial is complete as an act of power and empty as an act of explanation. It tells you that you lost without telling you anything you could use.

This is not a small discourtesy. A no that withholds its reasons withholds the very thing a person would need to do anything about it. You cannot correct an error you cannot see. You cannot supply the missing fact if you are not told a fact was missing. You cannot contest a judgment whose basis is hidden from you, and you cannot even decide whether contesting is worth your effort, because you have been given nothing on which to estimate your chances. The bare no does not merely fail to help. It actively forecloses every path the person might have taken, and it does so while appearing to be a neutral statement of fact.

What a good no contains

A good no has a shape, and the shape is not elaborate. It has four parts, and each one answers a question the person actually has. It states what was decided — plainly, without euphemism, so there is no ambiguity about whether this was a denial. It states on what basis — the actual grounds, the evidence and the rule that produced the outcome, not a generic gesture at unspecified criteria. It states what would have changed the answer — the counterfactual, the thing that, had it been different, would have flipped the decision, because that is what turns a verdict into something a person can respond to rather than merely endure. And it states where to go to contest it — a real and reachable path, not a dead reference to a process that does not answer.

Notice that none of these four is a concession. A good no is still a no. It does not soften the outcome or apologize for it or pretend the answer is other than what it is. It is, if anything, more decisive than the bare no, because it stands behind itself. It does not hide. The institution that issues a good no is saying: here is exactly what we decided and why, here is what it would have taken to decide otherwise, and here is how you may challenge us if you think we are wrong. That is not weakness. That is the posture of an institution that believes its decision can survive being looked at.

The bare no asserts that the answer is correct. The good no offers the means to find out. Only one of them is willing to be wrong.

Power and accountability

This is finally what separates the two. A bare no is an exercise of power in its purest form — the capacity to determine an outcome and impose it without owing anyone an account. It works because the person on the other end has no purchase on it: nothing to grip, nothing to question, nothing to push against. A good no is the same outcome rendered accountable. It converts the act of denial from something done to a person into something a person can examine, test, and answer. The decision is unchanged. What changes is whether the person has any standing in relation to it, and standing is the whole of the difference.

What a good no requires, practically, is that the decision was recorded well enough to be explained — and explained truthfully. The four parts cannot be improvised after the fact without becoming fiction, a flattering reconstruction of a reason that may not have been the operative one. They have to come from the decision itself: the basis must be the basis that was actually used, the counterfactual must follow from the rule that was actually applied, the contest path must lead somewhere that can retrieve what actually happened. A Decision Receipt that carries the evidence consulted, the rule in force, and enough state to replay the outcome is what makes a good no possible, because it lets the denial speak from the record rather than from a story composed to soothe. Without that record, even an institution that wants to issue good noes cannot; it has nothing honest to put in the four boxes.

So the next time you watch a system tell someone no — or the next time a system tells you — the question worth asking is not whether the answer was kind. It is whether the no had a shape. Did it say what was decided, on what basis, what would have changed it, and where to go to object? If it did, you are looking at a denial that is willing to be held to account. If it did not — if it was a bare no, complete and closed and answerable to no one — then what you are looking at is not a decision at all in any sense a person can engage with. It is power wearing the clothes of a decision, and the difference between those two things is the difference the record was meant to carry.

— Dispatches · Summit Cognitive