{"schema_version":1,"experiment_id":"eoa_inverse_innovation_exp04_retrieval_first_paired20_20260802","cell_id":"deadweight_loss_reduction__systems_cybernetics","considered_ids":["H2","H4"],"selected_id":"H4","dispositions":[{"hypothesis_id":"H2","lane":"REJECT","reason":"The critic found substantial prior-art overlap across risk-tiered scrutiny, selective validation, controller simulation, monitored rollout, and rollback, while the remaining end-to-end integration claim has weak problem support and depends on an unlocated outcome evaluation."},{"hypothesis_id":"H4","lane":"FINALIST","reason":"Despite substantial collision with existing deprecation practices, the critic isolates a narrow, falsifiable residual: adoption-time expiry for every normative interface requirement, with renewal contingent on a recurring combined test of dependency, migration cost, security exposure, and asymmetric producer-consumer tolerance. The underlying legacy-compatibility problem has strong direct-source support."}],"selection_reason":"H4 is the sole candidate with both strong support for a recognizable problem and a material, testable residual distinction. Selection is limited to its default-reversal and mandatory joint renewal test, not the already-established broader practice of telemetry-guided staged deprecation."}