The decision and its double

There is an old idea, older than any of the technologies we now argue about, that a thing of consequence should have a second self — a likeness made at the same time as the original, kept against the day the original is gone or contested. The wax seal pressed beside the signature, the carbon beneath the form, the duplicate filed in another building so that fire in one place could not erase the truth. The instinct is sound and very old: anything that matters enough to be questioned later should be made twice. What changed is that we stopped applying the instinct to decisions, exactly when decisions started being made faster than anyone could remember making them.

I want to take the idea seriously and give it a name. Every consequential decision should have a double — a record produced at the same instant as the decision, made of the same material, that can stand in for the decision when the decision itself is questioned. Not a description of the decision written afterward. Not a summary composed for the file. A double: something that came into being alongside the decision and carries enough of it that, later, when the decision is gone and the people who made it have moved on or forgotten, the double can answer in its place.

The distinction between a double and a description is the whole essay, so it is worth being precise about it. A description is made after, by someone looking back, and it inherits all the weaknesses of looking back: it remembers the parts that mattered to the person remembering, it smooths the parts that were messy, it knows how the story ended and quietly arranges the beginning to suit. A double is made during, before anyone knows whether the decision will be challenged, and so it has no motive to flatter. It is not a witness recalling the event. It is a piece of the event, set aside.

The test of a double is simple and unforgiving: can it answer for the decision without the decision present? If it cannot, it was never a double. It was a souvenir.

The test of the double

Here is how to know whether what you are holding is a double or merely something that resembles one. Imagine the decision itself has vanished — the moment is past, the system that made it has changed, the person who made it is unavailable or no longer remembers. Now someone with standing arrives and asks the decision to account for itself. The decision cannot answer; it is gone. So the question falls to the record. Can the record answer in its place? Can it produce what was actually known at the time, the rules that were actually in force, the path the decision actually took — fully enough that a careful outsider could replay it and reach the same place, or find precisely where it diverges?

If the record can do that, it is a double. It stands in for the decision because it carries the decision's substance, not its reputation. If it cannot — if all it offers is a confident paragraph about what was decided and a reassurance that the right factors were weighed — then it is not a double at all. It is a souvenir of the decision: a memento that proves the decision happened without preserving anything that would let you test it. A souvenir is pleasant to have and worth nothing in a dispute, because the one thing it cannot do is the one thing you need it to do — answer for the original when the original is absent.

This is a harder standard than it first appears, and most of what institutions keep fails it. Logs that record that a decision was made, but not the evidence it rested on. Files that name the policy but do not freeze its text. Notes that capture the conclusion and discard the reasoning. All of these are descriptions wearing the costume of evidence. They satisfy the impulse to keep a record without satisfying the purpose of keeping one. They will reassure an auditor who does not push, and they will collapse the instant someone pushes — which is the only moment a record is ever actually needed.

Why decoration fails exactly when it is needed

The failure has a cruel timing to it. A record that is decoration rather than evidence works perfectly right up until the moment of challenge, and then it fails completely. While no one is contesting the decision, the souvenir is indistinguishable from the double; both sit in the file looking official. The difference only surfaces under pressure, which is precisely when you can no longer go back and make a proper double, because the moment that would have produced it is gone. You cannot reconstruct, after a challenge, the evidence as it actually stood before the challenge — memory has already been reshaped by knowing the decision was disputed. The window to make a double is the same instant as the decision. Miss it, and no amount of diligence afterward can reopen it.

This is why a Decision Receipt has to be struck at the moment of the decision and not assembled later from recollection. Its whole claim to standing is that it is a double and not a description — that it came into being with the decision, carrying the real inputs and the frozen rules and enough state to be run again, and that it has had no opportunity since to be revised toward a more comfortable account. The receipt is admissible precisely because it could not have been edited to win the argument it is now being used to settle. It was made before the argument existed.

So the question to ask of any record that claims to account for a serious decision is not whether it looks complete. Souvenirs look complete; that is what they are for. The question is whether it could stand in for the decision with the decision gone — whether, asked to answer in the original's place, it could actually answer. A record that can is evidence, and it will hold when everything around it is in doubt. A record that cannot is decoration, and the day you reach for it will be the day you discover, too late, that you kept the likeness and lost the thing it was supposed to preserve.

— Dispatches · Summit Cognitive