Definition
The property that no single provider is necessary to both execute a consequential process and evaluate it. Execution and evaluation may be bought in the same market, but the institution must still be able to run the evaluation when the executing provider withdraws, degrades, or disputes the result.
A system that grades its own homework has no detection component, however much telemetry it emits, so detection is not satisfied by instrumentation the executing provider alone supplies or controls, and the evaluator has to be named. The property is carried by authority_grant.correction_capacity.detection.independent_of_executor and dependency_record.independent_evaluation, and bound by STD-08 §4.6.
It matters because detection is the first component of correction capacity, and a dependence that captures detection removes correction while leaving the diagram intact. Independence is the part of dependency state that substitution cost alone will not show, since an institution can afford to replace a provider and still be unable to check one.
Tests that show it is in place
- For each consequential process, the executing provider and the evaluating party are named separately, and the evaluator can report while the executor disputes the finding.
- Detection signals are reproducible from a source the executing provider does not supply or control.
- The evaluation has been run in a rehearsal with the executing provider's own telemetry withheld.
Evidence to keep
- Artifact: A grant record whose detection component names an evaluator independent of the executor, matched by the dependency record's independent_evaluation entry.
- Behavior: An evaluation completed and reported while the executing provider's telemetry was unavailable or contested.
- Metric: Share of consequential processes whose execution and evaluation do not depend on one provider.
References
- STD-08 delegation · Standard
- Dependence and reversibility eval · Eval