The half-life of a justification

We treat a justification as if it were a stone — set it once, at the moment of the decision, and it holds. The evidence was sufficient, the rule was applied, the box was ticked, and the reasoning is filed away as settled. But a justification is not a stone. It is closer to a battery. It is charged by the conditions that made it valid, and from the instant it is made it begins, slowly, to discharge. The evidence it rested on gets older. The population it generalized from drifts. The assumption it took for granted stops being true. None of this is visible from inside the decision, because the decision does not move. The world it was about does.

This is the thing standard practice is built to ignore. We are good at the moment of judgment — gathering the inputs, weighing them, producing an outcome we can defend on the day. We are far worse at the long tail afterward, the months and years in which that outcome continues to govern something while the justification underneath it silently weakens. A loan is still being serviced on a risk model trained on a world that no longer exists. A flag is still attached to a name on evidence that was thin when it was fresh and is now merely old. A policy still fires because no one has asked, lately, whether the reason it fires is still a reason.

A justification has a half-life. The question is never only whether it was sound when made, but how much of that soundness is left now — and whether anyone is responsible for measuring the decay.

Calling it a half-life is meant literally, not as a flourish. Decay is rarely a cliff; it is a curve. A justification does not stay fully valid and then fail all at once. It loses a little of its force with each shift in the world it described, and the rate of loss is itself a property worth knowing. Some justifications are nearly permanent — grounded in facts that do not move. Others are perishable, valid for a quarter, maybe a week. The dangerous ones are the perishable justifications treated as permanent: decisions that were genuinely well-made, cited for years afterward as if the ground beneath them had never shifted.

Decay you cannot see is decay you cannot answer for

The trouble is that a decision and its justification, once recorded, look exactly the same on the day they were made and a thousand days later. The file does not yellow. The confidence score does not fade. There is no visual signal that the account has weakened, which means the weakening has to be tracked deliberately or not at all. And "not at all" is the default. Most systems that can tell you what they decided and why cannot tell you when that why stopped being good enough — because they were built to capture justification as an event, not to maintain it as a state.

This is where the discipline gets specific. To take the half-life seriously, a justification has to carry more than its reasoning. It has to carry the conditions it depended on, marked as conditions — the freshness of its evidence, the assumptions it would not survive without, the window in which it can be trusted before it is due for review. A justification built this way can be asked, at any later moment, a question the stone version cannot answer: are you still good? And it can return something other than a confident yes by default.

Re-justifying is the work

The instinct, once you see the decay, is to want decisions that never need revisiting. That instinct is the error. The goal is not justifications that last forever; it is decisions that know when to be re-examined and institutions willing to do it. Standing behind a decision is not a thing you do once, at the moment you make it. It is a thing you keep doing, or stop doing, every day the decision remains in force. An organization that re-justifies its consequential decisions on a schedule it can defend is not being indecisive. It is being honest about the difference between a judgment that was right when made and one that is still right now.

The institutions that will be trusted as authority migrates into systems are not the ones with the most confident records. They are the ones that can show, when asked, that they have been watching their own justifications age — and acting before the decay became someone else's problem to discover. The alternative is to keep standing behind decisions whose reasons quietly expired, and to find out that they did only when an affected party, holding the original justification and a calendar, points out how old it has become.

— Dispatches · Summit Cognitive