Causal Provenance and the Epistemology of Artificial Minds: Construct Validity, Inferential Non-Transferability, and Measurement Dependence in Artificial-Mind Attribution

This paper develops an epistemological framework for attributing mental constructs to artificial systems. It distinguishes causal from epistemic provenance, separates operational criterion satisfaction from construct validity and ontological identification, and introduces a principle of inferential non-transferability across computational units. The framework is organised as a three-tier epistemic architecture and operationalised through a five-stage provenance protocol addressing conceptual independence, structural generalisation, causal-mechanistic intervention, evaluative perturbation, and construct-generating prediction. A hypothetical metacognition case illustrates how the protocol distinguishes causal mechanisms from representations of constructs and evaluative adaptations. The framework is theory-neutral regarding the ultimate ontology of mind and is intended to clarify the evidential conditions under which observations can legitimately support attribution of an artificial mental construct.