Accountability has a clock

We talk about accountability as though it were a fixed attribute of a decision, like its date or its author — something a decision either has or lacks, settled once and for all at the moment it is made. A decision was accountable, we say, or it was not, and we imagine we can return to that judgment at any later point and find it unchanged, waiting for us. This is a comforting picture and a false one. The capacity to account for a decision is not a property the decision keeps. It is a window that opens when the decision is made and begins, immediately, to close.

To account for a decision means to be able to show, later and to someone else, what it rested on: the evidence that was actually consulted, the rules that were in force, the state of the world the decision responded to. None of those things is preserved automatically. The evidence sits in systems that are constantly written over. The rules are revised, sometimes silently, sometimes with great ceremony, but either way replaced. The state of the world moves on. And human memory, which we lean on far more than we admit, is not a recording but a reconstruction that grows more confident and less accurate with every retelling. Each of these is a clock, and they are all running from the instant the decision is made.

Decay is the default

The mistake is to imagine that a record, once it exists, holds still. Most records do not. A log that points to a document does not protect the document; when the document is updated, the log now points to something that no longer says what it said. A note that an automated system was followed does not freeze the system; the next release changes the behavior, and the note, unchanged, now describes a process that no longer exists. What looks like a durable account is often just a durable pointer into a world that has kept moving. The account decays not because anyone destroys it but because everything it depends on quietly becomes something else.

This is why accountability after the fact so often fails, and fails in a particular, frustrating way. The question is asked in good faith — what did this decision rest on, and was that enough — and the answer that comes back is a reconstruction assembled from whatever survived. Today's version of the policy is quoted as though it governed yesterday's choice. The current state of a record is read back as though it were the state at the time. People recall, sincerely and wrongly, what they were thinking. None of this is fraud. It is the ordinary erosion of evidence under time, and it produces an account that is plausible, internally consistent, and not actually a record of what happened.

What you can prove about a decision is highest the instant it is made, and falls from there. Nothing you do later raises it; you can only slow the fall.

The window is at the moment

If decay is the default, then the practical consequence is sharp. The only moment at which a decision's real inputs are fully present is the moment of the decision itself. That is when the evidence consulted is the evidence as it actually was, when the rules in force are the rules as they actually stood, when the state of the world is the state the decision actually answered to. A fraction of a second later, the erosion has already begun. Preservation is not something you can defer and return to once the matter turns out to be important, because by the time it turns out to be important the materials you would need have already moved. The window to capture what a decision needs is open at the decision, and closing.

This reframes what a record is for. A record is not a courtesy you extend to the future when you happen to have time. It is a deliberate act of arresting decay at the one moment decay can be arrested — taking the evidence, the rules, and enough state to replay the decision, and fixing them in place before they become something else. A record built this way does not depend on later systems holding still, because it carried its dependencies inside itself. It is the difference between a photograph and a memory: one preserves the moment, the other reconstructs it under the pressure of everything that has happened since.

There is an uncomfortable corollary for how organizations behave. Because the cost of preservation is paid up front and the benefit is realized only later, if ever, the temptation is always to skip it. The decision worked out; no one is asking; why spend effort freezing a record nobody requested. But the requests, when they come, arrive on their own schedule, and by then the clock has run. You cannot reach back and capture what a decision rested on after the materials have turned over. The choice to be accountable is not made when the question is asked. It is made, or lost, at the moment of the decision, when almost no one feels the need.

So accountability is best understood not as a verdict you can always revisit but as a window you must act within. The discipline it asks for is unglamorous and badly timed: to preserve, at the moment of deciding, the things a future challenge will need, while the matter still seems routine and the cost feels unjustified. The clock is already running when you make the decision. What you save then is all you will ever have. Everything after that is reconstruction wearing the clothes of a record.

— Dispatches · Summit Cognitive