E.11:5.14 - A decision relies on a causal or counterfactual claim

A reader encounters “this policy would have prevented harm”, “the intervention caused the improvement”, “this fairness result is causal”, or “these simulated counterfactuals prove the decision”. Ask whether publication, choice, deployment, assurance, audit, benchmarking or method dispatch actually relies on causal support. Causal-looking wording alone is insufficient; the relied-on claim is the discriminating fact.

Open C.28 when the required support concerns an effect, intervention success, counterfactual comparison, causal fairness, policy optimality, causal evidence, off-policy/causal-RL evaluation or causal method superiority. Its Solution recovers the causal-use question, causality-ladder rung, claim kind, estimand, identification, counterfactual sampling realizability and support components, then distinguishes supported from unsupported use. A cheap downgrade sentence can be enough; a CausalUseTriageRecord, CausalUseEvidenceDesignRecord or CausalUseSupportResult is used only at the scope needed by the receiving claim.

If only a measured value is at issue, use C.16; if rate, trend or temporal adequacy is at issue, use C.27. If the question is method, work or work-plan structure, use A.15 and A.3.2; a sampling procedure, intervention recipe or target-trial phrase does not supply causal support by itself. Split a boundary duty or agreement with A.6. Use C.26 for residual QL modeling only when an exact contextual-model obstruction still changes inference or action after the ordinary subject assertion and predicate are in place.

A.10 governs the evidence/provenance paths actually relied on. B.3, D.5, G.5 and G.9 use a C.28 support result when the downstream question is respectively assurance, ethical/fairness audit, method dispatch or benchmark parity. Keep C.11, C.19 or C.24 as the direct local-choice, live-pool-policy or fixed-action call-planning use when that is the actual work; the support result does not decide it. Stop at the sufficient downgrade, scoped causal record/support result, or named neighboring use that cites its limits.

Return to causal support if the estimand, causal rung, identification, realizability or relied-on use changes. A randomized procedure is not automatically counterfactual support; simulation is not realized counterfactual data; a fairness metric is not causal fairness; and methods on different causal rungs or estimands are not made comparable by one benchmark. Causal support itself creates no duty, promise, commitment, release gate or admissibility predicate.