A.10:4.7 - Causal support in evidence-provenance paths
An A.10 path used for a causal claim cites the exact C.28 components it actually carries; it does not compress them into one alternative-valued “support basis” or copy C.28’s field list.
causalSupportComponentRefs?: CausalSupportComponentRefs
causalUseSupportResultRef?: CausalUseSupportResultRef
CausalSupportComponentRefs remains defined only by C.28. A.10’s ordinary evidence-provenance path still identifies the exact source, carrier, Work or data, provenance relations, and bounded use required by this pattern. Inside the C.28 contract, cite only the components the causal claim actually relies on. When C.28 admits another specialist component, A.10 can cite it through that contract without defining a second schema.
Examples:
- an observational cohort path cites the observation and measurement Work plus
observationalOrNaturalBehaviorData; an intervention-effect statement still needs a C.28 identification or design result; - a randomized estimate path cites assignment and Work evidence, its identification or design result, estimate, uncertainty, and limits;
- a prospective counterfactual-sampling path may cite the realizability result, its decision Method, the sampling construction or obstruction required by its status, and any unresolved question; counterfactual-quantity bounds use a separate identification result, and the prospective path claims no performed sampling or data;
- a performed counterfactual-sampling path cites independently admitted dated sampling Work and the resulting sample or data; add precise assignment-bound attribution only when the receiving support claim uses it. Only the complete Work/data path may support
realizedCounterfactualSamplingData; - a simulation path cites model output, assumptions, validation, and bounded model use; it does not become realized or interventional evidence by relabeling;
- a target-trial emulation path cites its
TargetTrialMappingResult, including the observational source, protocol-to-data mappings, gaps, residual-confounding assessment, and sensitivity mappings; reporting completeness alone establishes neither identification nor low bias; - an off-policy or causal-RL path cites its exact
OffPolicyCausalEvaluationResultwith the question, behaviour and evaluation policies, overlap check, supported and unsupported use, and reopen condition. It also cites optional details about history or horizon, confounding, changed endpoints or transport, the estimator, and uncertainty only when they change what the path supports, where that support came from, whether it is current, its bounded use, or when it reopens; - a causal-representation path cites its exact
CausalVariableRepresentationRecordwith the question, source representation, selection or abstraction Method, representation assumptions, intervention-validity result, supported and unsupported use, and reopen condition. It also cites optional invariance, fidelity, query-preservation, uncertainty, or shift-limit results only when they change what the path supports, where that support came from, whether it is current, its bounded use, or when it reopens; - a transport path cites every changed endpoint, assumptions, overlap evidence, formula or comparator, uncertainty or unresolved assumptions, and bounded use carried by its transportability result; and
- an observationally identified estimate cites both the evidence path and separate identification and estimate results.
What changes in practice: the path exposes where every relied-on component came from and may cite the C.28 support-result episteme. A.10 creates none of the C.28 components, the causal verdict, or downstream authority by carrying their refs.