Assertions produced by one system are routinely acted upon by another, and the receiving system acts on what a claim says rather than on what supports it. Where the originator may be a generator rather than an observer, a well-formed claim can acquire an authority its basis never warranted, with no falsehood introduced at any step. This dissertation names that failure False Determinism, the unsanctioned elevation of a claim's epistemic authority at a handover boundary in the absence of evidence sufficient to warrant it, and argues that the remedy is action-relative: whether a basis suffices is a relation between that basis and the act it would license, not a property of the claim.
The work develops that thesis into an executable discipline. A portable per-claim basis record carries how evidence was acquired, what operation produced the claim, a typed set of the events it descends from, its authority and validity, its establishment time and scope, and how each field of the record itself came to be populated. An inheritance rule governs what survives derivation: a derived claim's origins are the union of its evidential parents' origins, less any explicitly elided, plus fresh acquisitions, plus the typed event that produced it, which is what prevents a generated claim presenting as a plain calculation after a further step. An admissibility method takes a basis and the demand of a specific act and returns admit, refuse or undetermined with a trace naming the axis that decided. The artefact is a Python library with an executable specification written before the implementation.
Evaluation followed a sealed protocol. The artefact was frozen before any scenario existed. Thirty-nine scenarios were then authored from external sources, each recording its situation and its reference warrant separately, decomposed into atomic requirements with their logical relations, and sealed before any translation into the artefact's terms was attempted. Translations were sealed before execution, and only those classified as exactly faithful were eligible to be run.
The principal result is representational. Five translations were classified exact before execution; post-execution validity analysis identified one mistranslation, leaving four defensible exact translations from thirty-nine frozen requirements. On that same post-analysis view, ten were stricter than the requirement they encoded, five were weaker, and twenty could not be encoded at all — a partition that sums to the thirty-nine with the reclassified case counted among the stricter rather than the exact. All four defensibly exact cases were stratum 2, whose warrants were reached through an inferential bridge supplied by this work; no case whose warrant is stated outright by an external source was exactly representable. The largest identified deficiency is the discipline's own central move: the demand language cannot state that an available fact does not itself warrant a conclusion.
Discrimination was therefore measurable only over the eligible cases. Three of five matched on both verdict and predicted decisive axis as executed, and three of four among the translations surviving validity review, the latter reported as a sensitivity analysis rather than a result. One mismatch was an unwarranted admission arising where a derivation occurred outside anything the discipline mediated, so that no record carried the generated ancestry; the other was a defect in translation rather than in adjudication.
The result also identifies a further problem for increasingly automated decision loops. Where a probabilistic system replaces a human reviewer, its judgement becomes another generated claim, while the authority, independence and enforcement associated with the former human role do not transfer merely because the system performs the same task. This version does not address that case.
No broad discrimination claim follows from three cases, and no claim of operational usefulness follows from any of this. The contribution is a formally disciplined prototype, an executable account of its own boundaries, and evidence that property-based demands alone are insufficient for many externally grounded operational requirements. The falsification is specific: it concerns the representational adequacy of this version's demand language, not the existence of False Determinism, the action-relative thesis, or the correctness of the adjudication method on demands it can express.