{"schema_version":1,"experiment_id":"eoa_inverse_innovation_exp07_retrospective_selector60_20260803","cell_code":"E7C037","selector_replication":1,"assessments":[{"blind_id":"CANDIDATE_A","problem_reality_importance":87,"causal_archetype_fit":88,"distinctiveness_prior_art_resilience":58,"operational_specificity":89,"falsifiability_test_quality":85,"adopter_partner_path":81,"deployability_complexity":70,"authority_safety_reversibility":93,"strict_potential":77,"empirical_partner_potential":84,"scrutiny_priority":79,"biggest_visible_risk":"Control-relevant evidentiary meaning may remain inseparable from workbook formulas, protected ranges, signatures, or platform configuration, defeating behavioral substitution.","rationale":"The migration problem is consequential, the representation-independent contract is causally central, and the closed-case read-only replay is unusually bounded and decision-relevant. Explicit equations, role constraints, transition laws, typed failures, and successor history create a strong oracle. Scrutiny should test whether the abstraction genuinely captures the control rather than merely its normalized workflow."},{"blind_id":"CANDIDATE_B","problem_reality_importance":89,"causal_archetype_fit":78,"distinctiveness_prior_art_resilience":50,"operational_specificity":86,"falsifiability_test_quality":79,"adopter_partner_path":69,"deployability_complexity":54,"authority_safety_reversibility":90,"strict_potential":63,"empirical_partner_potential":73,"scrutiny_priority":67,"biggest_visible_risk":"Repository-native security, legal-hold, signature, and audit-log properties may not be reducible to black-box capsule behavior.","rationale":"Evidence custody is highly important and the proposal specifies provenance, lineage, holds, authorization, disposition, and failure semantics well. However, its decisive claims span legal, security, records, and audit authorities, while the fabricated-data study cannot establish production-grade custody equivalence. Irreversible disposition is safely excluded, but the path from synthetic conformance to adoption remains comparatively long."},{"blind_id":"CANDIDATE_C","problem_reality_importance":91,"causal_archetype_fit":94,"distinctiveness_prior_art_resilience":63,"operational_specificity":94,"falsifiability_test_quality":93,"adopter_partner_path":89,"deployability_complexity":80,"authority_safety_reversibility":94,"strict_potential":88,"empirical_partner_potential":92,"scrutiny_priority":91,"biggest_visible_risk":"The contract or reference model could encode the wrong allocation policy while still producing internally consistent, conserving results.","rationale":"This is the strongest bounded causal use of the archetype: formula representation directly causes hidden recipient-level differences, and the proposed contract exposes the policy-level semantics needed for substitution. Exact-arithmetic differential tests, predeclared outputs, metamorphic relations, recipient-level comparison, and a closed-period read-only pilot can decisively reject the intervention without posting anything. The principal scrutiny question is whether its semantic model remains adequate for real allocation methods, including nonlinear or reciprocal rules."},{"blind_id":"CANDIDATE_D","problem_reality_importance":90,"causal_archetype_fit":90,"distinctiveness_prior_art_resilience":55,"operational_specificity":90,"falsifiability_test_quality":88,"adopter_partner_path":86,"deployability_complexity":84,"authority_safety_reversibility":94,"strict_potential":82,"empirical_partner_potential":88,"scrutiny_priority":85,"biggest_visible_risk":"A clean population interface and passing oracle could conceal defects in source extraction or transformation controls that auditors must inspect directly.","rationale":"The exact-once population, scope, totals, sealing, lineage, pagination-failure, and immutability requirements form a concrete substitution test around a real ERP-migration problem. The closed-period adapter study is safe, feasible, and capable of changing the decision. Its contrastive claim is somewhat vulnerable because canonical journal models, reconciled extracts, and conformance testing look close to the proposed practice; scrutiny must determine whether the behavioral boundary adds meaningful protection beyond those approaches."}],"rank_order":["CANDIDATE_C","CANDIDATE_D","CANDIDATE_A","CANDIDATE_B"],"top_choice":"CANDIDATE_C","portfolio_observation":"All four proposals use the archetype substantively and offer reversible pilots, but they differ in how completely black-box behavior can capture the governed object. Allocation semantics are the most mathematically discriminating and partner-testable; journal populations and reconciliations retain increasing dependence on source controls or assurance artifacts; evidence custody has the widest gap between synthetic behavioral testing and production, legal, and security sufficiency.","confidence":"HIGH"}