Operative change
Type: kb/types/definition.md · Tags: foundations, self-improving-systems
A change is operative when it affects the system's subsequent operation over the relevant horizon: it persists long enough to matter there, and it has a path — a consumer, a channel, and a force, in behavioral authority terms — through which it actually reaches later behavior. It is the self-improving system definition's guard against changes that exist without consequence.
Scope
Two conditions, both needed:
- Persistence over the relevant horizon. The horizon is declared, not fixed by the definition. Cross-episode persistence is the usual reading for learning systems, but a shorter horizon can be the relevant one: a plan revised mid-deployment that governs the rest of the deployment window is operative for that window. Operativity does not require permanence — many useful adaptations are temporary and still operative, and a change later rolled back was operative while it held.
- A behavioral-authority path. Something must consume the change, through some channel, with some force. A reviewed note nothing loads, an approved patch never merged, a generated validator nothing invokes — each persists and affects nothing.
Exclusions
- Transient compensation. An adjustment consumed entirely within the moment that produced it — the eye adapting to the dark, a control output — determines nothing afterward.
- Inert persistence. An artifact that survives without a consumer is stored, not operative. Persistence alone is the cheap half.
- Permanence as a requirement. Demanding that a change last forever, or survive retraining, confuses operativity with immutability; the definition asks only that the change operate over the horizon that matters for the claim being made.
Misuse Cases
- Calling a change operative because it was accepted or written down, without naming what consumes it.
- Denying operativity to a temporary adaptation that demonstrably steered behavior over its window.
- Reporting operativity without the horizon: "operative" with no stated horizon is as underspecified as an autonomy grade with no declared boundary.
Relevant Notes:
- Self-improving system — defined-in: the definition whose "operative changes" clause this term sharpens
- Behavioral authority — grounds: the consumer, channel, and force that constitute the path to later behavior
- Behavior-determining organization — contrasts: what the change must land in; operativity asks whether it then takes effect