Every decision has a second reader

Every decision is written for a first reader, and almost every decision forgets the second. The first reader is the person making the call. They have the context already; they were there when the situation developed; they hold in their head the half-dozen things that did not get written down because, at the time, they did not need to be. To this reader the record can be terse, allusive, full of shorthand. A note that says approved per usual criteria is perfectly legible to the person who knows what the usual criteria are and why they applied today. The first reader and the writer are, in effect, the same mind a moment apart. They share everything.

The second reader shares nothing. The second reader arrives later — sometimes much later — and has to evaluate the decision without having been present when it was made. This is the auditor reconstructing what happened, the court asked whether the call was lawful, the successor who inherited the file, the affected party who wants to know why the answer was no. They do not have the context. They were not there. They do not hold the half-dozen unwritten things, and they have no way of recovering them except from what the record chose to preserve. To the second reader, approved per usual criteria is not a justification. It is a locked door.

The whole difficulty of accountable decision-making lives in the gap between these two readers. A record that satisfies the first reader is trivial to produce, because the first reader needs almost nothing. A record that satisfies the second reader is hard, because the second reader needs nearly everything — and the writer, sharing the first reader's mind, can barely perceive what is missing. We write what we would need to remember. We do not write what a stranger would need to understand, because we are not strangers to ourselves.

You are not writing the record for the version of you that already knows. You are writing it for the stranger who arrives after the context is gone.

The second reader does not share your assumptions

What makes the second reader so demanding is not hostility. It is unfamiliarity. The second reader does not know which sources you trusted and which you discounted, so the record has to say. They do not know which rules were in force on the day, because rules change and the second reader is reading on a different day, so the record has to carry the rules as they stood. They do not know the order in which the evidence arrived, and order shapes meaning, so the record has to preserve the sequence rather than the summary. None of this is exotic. It is simply the difference between a note you write to yourself and a letter you write to someone you have never met about a situation they did not witness.

This is also where automation changes the stakes rather than inventing them. A human decision-maker can, in principle, be found and asked — the second reader can knock on a door and say, walk me through this. A system cannot be asked. When the question comes years later, the people are gone, the context has evaporated, and the only thing standing between the decision and the second reader is whatever the record held still. So the record is not a courtesy that supplements a living account. It is the entire account, and it has to be built for someone who will read it cold.

Designing for the second reader changes what you record, concretely and immediately. You stop recording conclusions and start recording grounds, because the second reader cannot evaluate a conclusion they cannot trace. You stop pointing to policies and start carrying them, because a pointer resolves to whatever the policy says now, not what it said then. You stop trusting that the obvious will stay obvious, because the obvious is exactly what evaporates first. The discipline is not to write more. It is to write for absence — to assume that everything you currently take for granted will be unavailable to the person who matters most.

Convenient for the first reader, defensible for the second

There is a real cost to writing for the second reader, and it is paid by the first. It is faster, cleaner, and more comfortable to record only what you need. The terse note, the shorthand, the conclusion without its grounds — these are conveniences, and they are conveniences precisely because they serve the reader who needs the least. A decision optimized for the first reader is convenient. A decision optimized for the second reader is defensible. The two rarely look the same, and when an organization is honest about which one it is producing, it is usually producing the first while calling it the second.

This is what standing comes down to, in the end. The affected party who wants to contest a decision is a second reader with the highest possible stakes and the least possible context. Giving them a record they can actually read — sources, rules, sequence, enough to reconstruct rather than merely to be told — is what converts a right to contest into a real ability to contest. A record written only for the first reader hands the second reader a locked door and calls it transparency. The asymmetry is the whole problem, and the only thing that closes it is a writer who, at the moment of decision, takes the second reader seriously enough to write for someone who is not there.

So the test is not whether the record reads well to you. Of course it does; you are the first reader, and you already know. The test is whether it would read to a stranger who arrives after you are gone, who shares none of your assumptions, and who has every reason to doubt. Write for that reader, and you have a decision you can defend. Write for yourself, and you have a decision that was only ever convenient — and convenience, when the second reader finally arrives, is not a defense.

— Dispatches · Summit Cognitive