A judgment you are not recomputing is a memory

Entry XXVI makes a claim about rationality that most people accept in principle and almost nobody builds for. Rationality is not a permanent condition. It is maintained against constant erosion, and left unattended, judgment yields to habit and to unexamined routine.

The line that does the work: clarity does not sustain itself. Knowing what is correct does not execute it.

That is usually read as being about effort. Keep paying attention, keep checking yourself. But there is a harder version, and it is structural rather than moral.

The harder version

A conclusion is correct at the moment you reach it. Then the inputs move.

If nothing forces you to recompute, the conclusion stays where it is, displayed at full confidence, indistinguishable from a live judgment. It does not announce that it has aged. There is no symptom. It looks exactly like the answer you would arrive at today, because it is written in the same place, in the same words, with the same authority.

That is drift, and it is worse than the kind XXVI describes, because the kind you can feel at least gives you something to resist. A cached verdict gives you nothing to notice.

So the test is not whether you still believe a conclusion. It is whether anything in your arrangement would tell you if it had stopped being true.

What happened

I keep a scoring instrument for a decision I am making slowly. It ranks candidates against criteria, produces a weighted score, and displays a rank beside each one.

This week I corrected a fault in the weighting. Before publishing the correction, I checked what the board had actually been saying.

The displayed rank was a fixed label. The score beside it was a live calculation. Nothing connected them. At some point the scoring had changed, the scores had moved, and the labels had stayed exactly where they were.

So the board showed my leading candidate in third place. Recomputed from its own current scores, it was fourth. The label had been wrong for some time and had reported nothing, because a number typed into a cell does not know it has been superseded.

Then the correction landed, and the corrected weighting put that candidate back in third.

Why that is the uncomfortable part

The published answer was right. It had been right all along.

Which means that if I had not looked at the mechanism, I would have had every reason not to. The output agreed with itself. Correcting the weighting changed nothing I had told anyone. A person checking my conclusion against my method would have found them consistent.

The agreement was luck. A stale label and a corrected calculation happened to land on the same number, and the coincidence was doing the work of a verification I had not actually run.

That is the failure mode worth naming. Not being wrong, which announces itself eventually. Being right by a route that has stopped working, which does not.

The practice

The operational form:

For any conclusion you display, ask what would happen if its inputs changed. If the answer is that it would keep displaying the old value, it is not a judgment. It is a memory with a judgment's formatting.

Then either make it recompute, or make the disagreement visible. A frozen figure is legitimate when it is deliberately frozen and flagged as such. It is not legitimate when it is simply a live figure that stopped being updated.

The same question applies outside spreadsheets. A view you hold about a colleague, a market, a city, a person's reliability. Formed on evidence, correct at the time, and displayed since without recomputation. Nothing in most lives forces the recalculation, which is why the views that go stale are rarely the ones anyone argues about.

The standard

Drift is not only the erosion of attention. It is also the persistence of a conclusion past the conditions that produced it.

Rationality survives through repeated correction, not through insight alone. A conclusion that cannot be shown to be current is not being held. It is being remembered.

https://hoc-est-opus.com/leadership-standard/vigilance-and-rest

Free to read. Subscribing is optional.