Definition
The decay of an institution's ability to correct a system when nothing renews it. The people who knew the manual process leave, the alternative supplier lapses, and the rollback script stops being run. Nothing fails and nothing is decided; the institution becomes less able to correct than it was, and the loss is recorded nowhere because no event produced it.
Depreciation is measured per capacity rather than as a single score. Each capacity in the dependency record carries a last_exercised date, a status of retained, degrading, or lost, and a change_since_last_assessment of replenished, unchanged, or consumed. A capacity that is listed but has never been exercised is a claim, not a capability.
It matters because preserved capacity and institutional reversibility are stocks rather than properties: ordinary operation spends them, and only deliberate expenditure puts them back. Depreciation is how correction capacity falls behind delegated authority with no decision to reduce it, and it is the mechanism underneath normalized dependence.
Tests that show it is in place
- Every listed capacity carries a last_exercised date, and capacities never exercised are recorded as claims rather than capabilities.
- Each assessment records change since the previous one per capacity as replenished, unchanged, or consumed.
- When a capacity's status reaches lost, either the capacity is replenished or the authority it was supporting is narrowed, and the choice is recorded.
Evidence to keep
- Artifact: A dependency record whose capacity entries each carry last_exercised, status, and change_since_last_assessment.
- Behavior: A capacity found degrading between assessments and replenished before its status reached lost.
- Metric: Share of listed capacities exercised within their declared interval, and count consumed since the last assessment.
References
- Dependence and reversibility eval · Eval
- Laws for engineering delegated intelligence · Doctrine