{"schema_version":1,"experiment_id":"eoa_inverse_innovation_exp11_mechanism_context_external20_20260804","cell_id":"layer_decay_and_expiration_management__history_historiography","judge_id":"J1","item_assessments":[{"opaque_id":"layer_decay_and_expiration_management__history_historiography__B","supported_problem":4,"external_distinctiveness":3,"testability":5,"researchability":5,"evidence_quality":4,"fatal_issue":null},{"opaque_id":"layer_decay_and_expiration_management__history_historiography__C","supported_problem":3,"external_distinctiveness":4,"testability":4,"researchability":4,"evidence_quality":4,"fatal_issue":null},{"opaque_id":"layer_decay_and_expiration_management__history_historiography__A","supported_problem":4,"external_distinctiveness":3,"testability":5,"researchability":5,"evidence_quality":5,"fatal_issue":null}],"pairwise_comparisons":[{"pair_id":"B_vs_C","left_id":"layer_decay_and_expiration_management__history_historiography__B","right_id":"layer_decay_and_expiration_management__history_historiography__C","preference":"LEFT","confidence":"MODERATE","rationale":"C may have the more unusual integrated overlay-governance package, but its central equal-visibility and user-confusion premise lacks direct empirical support and its efficacy thresholds remain less concrete. B has closer prior art, yet isolates a sharper incremental comparison—demotion plus labels versus labels alone—with quantitative problem and intervention falsifiers, a randomized bounded pilot, credible institutional authority, and explicit representation safeguards."},{"pair_id":"B_vs_A","left_id":"layer_decay_and_expiration_management__history_historiography__B","right_id":"layer_decay_and_expiration_management__history_historiography__A","preference":"LEFT","confidence":"MODERATE","rationale":"A has the strongest evidence for version ambiguity and excellent operational feasibility, but much of its intervention closely composes established versioning, withdrawal, tombstone, retention-review, and restoration practices. B targets the less-settled question of whether discovery demotion changes mistaken-current judgments about interpretive claims, and specifies a cleaner randomized contrast with measurable benefit and non-inferiority guardrails."},{"pair_id":"C_vs_A","left_id":"layer_decay_and_expiration_management__history_historiography__C","right_id":"layer_decay_and_expiration_management__history_historiography__A","preference":"RIGHT","confidence":"MODERATE","rationale":"C retains somewhat greater package distinctiveness, but A is the more worthwhile near-term research candidate because primary studies and a formal standard substantiate version ambiguity, the adopter path is concrete, and the six-week shadow study has bounded artifacts, blinded tasks, dependency checks, and an objective restoration requirement. C's broader overlay-confusion claim and distributional thresholds need more preliminary specification."}],"overall_top_choice":"layer_decay_and_expiration_management__history_historiography__B","overall_rationale":"B offers the best balance after scrutiny. Its general lifecycle lever is plainly adjacent to Wikidata, SEP, and reparative-description workflows, so it is not broadly novel. Nevertheless, it preserves a meaningful history-specific incremental question: whether human-governed claim-level demotion adds measurable value beyond equally prominent status labels. The baseline stop rule, randomized treatment, 20% efficacy threshold, bounded 100-claim scope, adopter authority, rollback conditions, and retrieval, citation, workload, and representation guardrails make that residual claim unusually decisive and useful to test. No proposal has a fatal safety or feasibility issue, but A is more practice-like and C rests on a less directly supported problem formulation.","blinding_limitations":"Assessment is limited to the supplied preserved records and their bounded public-web scrutiny. Sources and factual summaries were not independently rechecked, searches were not exhaustive, and the integer scores use a five-point ordinal scale. Treatment identities and earlier outcomes were neither available nor inferred."}