{"schema_version":1,"experiment_id":"eoa_inverse_innovation_exp04_retrieval_first_paired20_20260802","cell_id":"invariant_mode_decomposition_design__political_science","considered_ids":["H1","H3","H4"],"selected_id":"H4","dispositions":[{"hypothesis_id":"H1","lane":"RESEARCH","reason":"The residual comparison is falsifiable, but problem support is only moderate and the direct sources do not establish access to grievance, concession, or counterfactual intervention data needed for the proposed treatment."},{"hypothesis_id":"H3","lane":"COLLISION","reason":"The Acertas filing substantially discloses the multivariate political-event, VAR, spectral-stability, warning, and mitigation workflow; the remaining crisis-day validation protocol is too narrow and benchmark-dependent to constitute a material differentiated treatment claim."},{"hypothesis_id":"H4","lane":"FINALIST","reason":"Official oversight sources directly establish consequential referral-management failures, while the residual claim is narrow and prospectively falsifiable: whether modal-perturbation rankings predict unresolved or repeatedly transferred cases better than complaint volume, degree centrality, and existing risk scores."}],"selection_reason":"H4 offers the strongest combination of a directly documented operational problem and a recognizable, bounded treatment distinction. Its novelty is confined to the referral-eigenmode perturbation ranking and temporal holdout advantage, rather than the already established components of complaint networks, key-player analysis, and risk-based audit planning."}