Evidence-Grounded Closure in EXPERAND Dynamics

Disappearance of an unresolved-state marker does not establish that the condition it represents has been resolved. This specification separates five elements that must remain distinct: a candidate state update, the substantive target of resolution, the available evidence, an evidence-relative verdict, and the authorization of the update itself. A possible-world semantics yields three verdicts—CERTIFY, DENY, and HOLD—relative to a fixed resolution question. A separate binding check relates any certificate to the actual condition and procedure. Under the stated assumptions, the construction admits a maximal sound classifier, an observational indistinguishability obstruction, a pathwise non-erasure invariant, and stability of three-way decisions under nonempty nested evidence refinement. Recovery without return alone forces every representation of a condition to distinguish at least three mutually exclusive classes: Never-Held, Held, and Resolved. With the further freely chosen axioms that Resolved is entered only by certified removal from Held and that the status coordinate never decreases, non-erasure makes the terminal status a complete audit trail. Finite checks on the original 24-state EXPERAND witness, whose status coordinate is a minimal carrier of this ternary distinction and whose generators satisfy both axioms, reproduce eight illustrative worlds, five observation fibers, eighty proposed removals (eight accepted, seventy-two rejected), eight recovery-without-return backgrounds, and 6,305 nonempty nested evidence-set pairs with no CERTIFY-to-DENY or DENY-to-CERTIFY reversal under nested refinement. These results establish internal correctness under the stipulated model. They do not establish clinically validated detection of healing or phenomenal ownership. The generic guard and verdict mechanisms are standard; the forcing results are the EXPERAND-specific contribution.