{"schema_version":1,"experiment_id":"eoa_inverse_innovation_exp04_retrieval_first_paired20_20260802","cell_id":"invariant_mode_decomposition_design__political_science","arm":"RETRIEVAL_FIRST","round_index":0,"hypotheses":[{"hypothesis_id":"H1","title":"Coalition defection-mode targeting","problem":"Whips target visibly undecided legislators while coordinated combinations of grievances drive coalition collapse.","affected_stakeholder":"Legislative coalition managers","workflow_boundary":"From pre-vote member signals to allocation of negotiation effort","failure_mode":"Member-by-member scoring misses a weakly damped collective defection mode.","unit_of_analysis":"Legislator–vote episode","causal_lever":"Direct concessions and outreach toward the grievance combination with greatest estimated outcome sensitivity.","archetype_mapping":"Decompose an estimated vote-transition operator, classify persistent or growing modes, and perturb them against passage probability.","expected_value":"Earlier, more efficient prevention of coordinated defections without assuming the largest bloc is most pivotal.","falsifiable_claim":"On held-out close votes, mode-targeted outreach recommendations predict or avert coalition defections better than marginal legislator-level scores.","diversity_rationale":"Targets collective legislative dynamics, uses an explicit transition operator, and intervenes before a discrete vote.","mechanism_slugs":["eigendecomposition_workflow","modal_stability_analysis","modal_sensitivity_sweep"],"search_questions":["Has modal or eigenvalue analysis been used to forecast legislative coalition breakdown?","Do whip systems optimize outreach over combinations of member grievances rather than individual probabilities?","What vote-history data can identify stable collective defection modes?"]},{"hypothesis_id":"H2","title":"Residual audit for consultation compression","problem":"Policy-consultation summaries can erase rare but consequential objections while retaining high-frequency themes.","affected_stakeholder":"Small or politically marginalized respondent groups","workflow_boundary":"From consultation submissions through coding and compression to the official briefing","failure_mode":"Variance-ranked theme reduction treats low-volume structured concerns as disposable noise.","unit_of_analysis":"Submission–theme pair","causal_lever":"Require residual reconstruction by respondent group and restore themes whose omitted residual is structured or consequence-heavy.","archetype_mapping":"Build a low-dimensional theme basis, reconstruct submissions, and govern retention with disaggregated residual thresholds and interpretation limits.","expected_value":"More compact briefings that preserve politically consequential minority information.","falsifiable_claim":"Compared with frequency-based summaries at equal length, residual-audited summaries retain more blinded-expert-identified consequential minority concerns.","diversity_rationale":"Addresses representational loss in administrative text compression, uses submissions rather than political actors, and changes a documentation gate.","mechanism_slugs":["principal_component_analysis","residual_reconstruction_test","spectral_decomposition_report"],"search_questions":["Are reconstruction residuals used to audit political consultation summaries?","How is minority-view loss currently measured in public-comment coding?","Which outcome labels can distinguish consequential rare themes from idiosyncratic noise?"]},{"hypothesis_id":"H3","title":"Escalation-mode early warning","problem":"Crisis dashboards track separate hostile actions but may miss a growing combination that makes escalation self-reinforcing.","affected_stakeholder":"Diplomatic crisis-management teams and exposed civilians","workflow_boundary":"From daily event coding to escalation-control decisions","failure_mode":"Stable averages conceal an unstable joint mode across rhetoric, mobilization, sanctions, and incidents.","unit_of_analysis":"Crisis-day","causal_lever":"Trigger de-escalatory measures when a locally estimated joint mode crosses a growth threshold, rather than when any single indicator peaks.","archetype_mapping":"Linearize short-window crisis transitions, partition modes by stability, and bind warnings to a declared local-validity window.","expected_value":"Lead time for intervention before surface indicators independently reach alarm levels.","falsifiable_claim":"Across historical holdout crises, unstable-mode alerts precede escalation transitions with better calibrated precision-recall than matched univariate thresholds.","diversity_rationale":"Uses temporal interstate crisis states, focuses on early warning rather than allocation or representation, and tests dynamic stability.","mechanism_slugs":["modal_stability_analysis","spectral_gap_monitor","eigendecomposition_workflow"],"search_questions":["Has local spectral stability been tested on international-crisis event streams?","Which crisis indicators form sufficiently stationary short-window state vectors?","Does modal warning add lead time beyond standard event-count or hidden-state models?"]},{"hypothesis_id":"H4","title":"Accountability-network intervention map","problem":"Oversight bodies prioritize agencies with many visible complaints while mutually reinforcing referral and responsibility structures diffuse accountability elsewhere.","affected_stakeholder":"Legislative and independent oversight offices","workflow_boundary":"From complaint and referral records to audit-target selection","failure_mode":"Raw volume rankings miss offices occupying the dominant mode of responsibility transfer.","unit_of_analysis":"Agency-office–reporting-period","causal_lever":"Prioritize audits or reporting-rule changes at nodes whose perturbation most reduces the network’s dominant responsibility-diffusion mode.","archetype_mapping":"Treat interoffice referrals as a connectivity operator, rank modal participation, then sensitivity-test candidate node interventions.","expected_value":"Audit effort aimed at structural accountability bottlenecks rather than merely busy offices.","falsifiable_claim":"Mode-and-sensitivity rankings identify future unresolved or repeatedly transferred cases better than complaint volume, degree, and conventional risk scores.","diversity_rationale":"Uses an organizational referral network, targets oversight portfolios, and evaluates node intervention against case-resolution outcomes.","mechanism_slugs":["network_spectral_centrality_analysis","power_iteration_probe","modal_sensitivity_sweep"],"search_questions":["Is spectral centrality already used to select public-sector accountability audits?","Which referral records permit construction of a responsibility-transfer network?","Does dominant-mode participation predict unresolved cases beyond degree and complaint volume?"]},{"hypothesis_id":"H5","title":"Critical-juncture model validity alarm","problem":"Institutional-reform teams reuse simplified scenario models precisely when a political shock changes which coalitions and veto points dominate.","affected_stakeholder":"Constitutional and electoral-reform planners","workflow_boundary":"From scenario-model maintenance to authorization of post-shock reform advice","failure_mode":"A reduced model remains operational after its retained modes rotate or lose separation from omitted modes.","unit_of_analysis":"Institutional configuration–scenario run","causal_lever":"Suspend or expand the scenario model when spectral-gap, mode-drift, or reconstruction-residual limits fail.","archetype_mapping":"Project reform dynamics into a reduced modal model, continuously test the retained-mode gap and residuals, and retire the model outside its scope contract.","expected_value":"Fewer confidently wrong recommendations at critical junctures while preserving fast scenario exploration in stable regimes.","falsifiable_claim":"A modal-validity gate rejects post-shock forecasts with materially higher error more often than calendar-based revalidation, while retaining comparably accurate forecasts.","diversity_rationale":"Governs model use at institutional regime shifts, treats scenario runs as units, and intervenes by suspending an analytic artifact rather than targeting actors.","mechanism_slugs":["reduced_order_model","spectral_gap_monitor","residual_reconstruction_test","spectral_decomposition_report"],"search_questions":["How are political institutional-scenario models currently invalidated after regime shifts?","Have spectral-gap or mode-rotation tests been used as political-model governance gates?","Which historical reforms provide pre-shock and post-shock scenario-validation data?"]}]}