{"schema_version":1,"experiment_id":"eoa_inverse_innovation_exp03_full320_20260801","cell_id":"invariant_mode_decomposition_design__sociology_anthropology","trajectory_id":"R","attempt_index":0,"archetype_slug":"invariant_mode_decomposition_design","domain_slug":"sociology_anthropology","decision":"CANDIDATE","problem_id":"neighborhood_collective_efficacy_erosion_blind_spots","causal_lever_id":"coupled_social_state_mode_intervention","proposal":{"problem":"HYPOTHESIS: Neighborhood collective efficacy can erode through coupled changes in trust, willingness to intervene, cross-group contact, institutional responsiveness, and visible disorder while ordinary indicator-by-indicator dashboards remain stable. Late recognition can permit avoidable escalation of withdrawal, disorder, conflict, and coercive response.","actors_substrate":["Residents and informal community leaders","Neighborhood associations and service organizations","Municipal outreach, public-space, and violence-prevention teams","Repeated resident surveys, service-response records, observational measures, and aggregated incident data","A resident-governed neighborhood-by-period panel dataset"],"observable_state":"For each neighborhood and period, a preregistered state vector records aggregated trust, expected mutual intervention, cross-group interaction, participation, perceived legitimacy, service responsiveness, visible disorder, and conflict. The observable signature is a repeatable multivariate direction that grows or weakly decays before adverse outcomes even when individual indicators remain below ordinary alert thresholds.","consequence":"Delayed or misdirected prevention, potentially followed by increased conflict, unequal surveillance, coercive intervention, and further loss of institutional legitimacy.","affected_objective":"Detect and reverse actionable erosion of collective efficacy early while minimizing stigmatization, surveillance, and coercion.","structural_mapping":[{"archetype_element":"A repeated transformation acts on coupled variables","domain_realization":"A locally estimated period-to-period transition maps each neighborhood social-state vector to its subsequent state.","claim_kind":"HYPOTHESIS"},{"archetype_element":"Invariant or approximately invariant directions","domain_realization":"Recurring combinations of trust, intervention expectations, contact, responsiveness, disorder, and conflict are treated as provisional social-state modes, not natural kinds.","claim_kind":"HYPOTHESIS"},{"archetype_element":"Scalar modal response","domain_realization":"Estimated gains distinguish combinations that decay, persist, oscillate, or amplify within the fitted observation window.","claim_kind":"INFERENCE"},{"archetype_element":"Action-relevant mode selection","domain_realization":"Modes are prioritized by out-of-sample outcome sensitivity, harm relevance, stability, and reconstruction contribution rather than variance alone.","claim_kind":"INFERENCE"},{"archetype_element":"Intervention in modal coordinates","domain_realization":"Resident-selected outreach, shared-space repair, facilitated contact, and accountable service-response changes are combined according to their estimated effects on the concerning mode.","claim_kind":"HYPOTHESIS"},{"archetype_element":"Residual, drift, and scope controls","domain_realization":"Unexplained subgroup patterns, changing mode directions, weak spectral separation, and regime shifts suspend modal decisions.","claim_kind":"INFERENCE"}],"component_map":[{"component":"Transformation Scope","status":"adapted","domain_realization":"A neighborhood-level transition operator estimated for a fixed cadence, geography, measurement protocol, and policy regime."},{"component":"State-Vector Definition","status":"adapted","domain_realization":"Aggregated, consent-governed measures of trust, mutual intervention, interaction, participation, legitimacy, responsiveness, disorder, and conflict."},{"component":"Invariant Mode Basis","status":"adapted","domain_realization":"Approximate eigen-directions of the fitted local transition, traced back to indicator and subgroup loadings."},{"component":"Modal Gain Spectrum","status":"direct","domain_realization":"Estimated eigenvalues with uncertainty intervals and non-normality diagnostics."},{"component":"Dominant Mode Selection Rule","status":"adapted","domain_realization":"Retain modes exceeding preregistered outcome-sensitivity or harm thresholds and contributing necessary out-of-sample reconstruction."},{"component":"Stable/Unstable Mode Partition","status":"direct","domain_realization":"Classify modes as decaying, marginal, growing, or oscillatory only within the stated local window."},{"component":"Modal Intervention Map","status":"adapted","domain_realization":"Map voluntary, resident-approved service and relationship-building actions to predicted changes in modal coordinates."},{"component":"Reconstruction Residual Check","status":"direct","domain_realization":"Reconstruct held-out social states and inspect total error plus structured subgroup and neighborhood residuals."},{"component":"Mode Drift Monitor","status":"direct","domain_realization":"Track mode rotation, reordering, gain changes, and residual growth at each refresh."},{"component":"Interpretation Scope Contract","status":"adapted","domain_realization":"Modes are provisional descriptive dynamics, not fixed cultural traits, individual propensities, or proof of causation."},{"component":"Mode-Coupling Register","status":"direct","domain_realization":"Record near-degeneracy, non-orthogonality, cross-effects, measurement overlap, and intervention spillovers."},{"component":"Local Linearization Window","status":"adapted","domain_realization":"Validity is limited to observed indicator ranges, participating neighborhoods, fixed cadence, and absence of major shocks or policy changes."},{"component":"Spectral Gap Threshold","status":"direct","domain_realization":"A preregistered minimum separation and bootstrap stability requirement gates reduced-mode use."}],"mechanism_dispositions":[{"slug":"eigendecomposition_workflow","disposition":"selected_load_bearing","contribution_type":"CORE_CAUSAL","adaptation_or_rejection":"Applied to the explicitly fitted local transition operator; conditioning and uncertainty must be reported.","counterfactual_removal":"Removing it eliminates invariant temporal directions and their gains, reducing the proposal to correlated indicators."},{"slug":"modal_sensitivity_sweep","disposition":"selected_load_bearing","contribution_type":"CORE_CAUSAL","adaptation_or_rejection":"Uses simulated and pilot-estimated perturbations to rank modes and action bundles by outcome leverage and spillover.","counterfactual_removal":"Removing it leaves no defensible link from descriptive modes to intervention choice."},{"slug":"modal_stability_analysis","disposition":"selected_load_bearing","contribution_type":"CORE_CAUSAL","adaptation_or_rejection":"Classifies fitted gains only inside the measured local regime and carries uncertainty around the stability boundary.","counterfactual_removal":"Removing it prevents identification of growing or weakly damped erosion modes."},{"slug":"mode_shape_testing","disposition":"considered_rejected","contribution_type":"NONE","adaptation_or_rejection":"Deliberately exciting a community as though it were a physical object is unsafe and natural shocks are confounded; held-out observation and bounded pilots replace it.","counterfactual_removal":"No material change because empirical validation is supplied by residual testing and the bounded pilot."},{"slug":"network_spectral_centrality_analysis","disposition":"considered_rejected","contribution_type":"NONE","adaptation_or_rejection":"Node centrality ranks actors, whereas the target is a coupled transition among social-state measures; it also risks stigmatizing central residents.","counterfactual_removal":"No material change to the state-transition causal chain."},{"slug":"power_iteration_probe","disposition":"considered_rejected","contribution_type":"NONE","adaptation_or_rejection":"Dominant-only estimation is inadequate because harmful low-amplitude modes and near-degenerate modes must remain visible.","counterfactual_removal":"No material change because full fitted-operator decomposition is feasible in the bounded test."},{"slug":"principal_component_analysis","disposition":"considered_rejected","contribution_type":"NONE","adaptation_or_rejection":"Variance directions may initialize diagnostics but do not represent period-to-period dynamics or intervention leverage.","counterfactual_removal":"No material change; covariance variance is not the proposed causal lever."},{"slug":"reduced_order_model","disposition":"selected_supporting","contribution_type":"OPERATIONAL","adaptation_or_rejection":"A small, explicitly local surrogate supports preregistered action-bundle simulations without being treated as a community digital twin.","counterfactual_removal":"The causal hypothesis remains, but bounded comparison of candidate bundles becomes slower and less transparent."},{"slug":"residual_reconstruction_test","disposition":"selected_load_bearing","contribution_type":"SAFETY_GUARDRAIL","adaptation_or_rejection":"Tests held-out reconstruction and searches residuals for omitted subgroup harms or nonmodal structure.","counterfactual_removal":"Removing it permits false compression and can hide socially consequential low-variance behavior."},{"slug":"singular_value_decomposition","disposition":"selected_supporting","contribution_type":"TEST_DESIGN","adaptation_or_rejection":"Used to diagnose non-normal amplification, conditioning, and low-rank robustness, not to relabel singular vectors as invariant social modes.","counterfactual_removal":"The main chain remains, but fragile eigenmode interpretations are less likely to be detected."},{"slug":"spectral_decomposition_report","disposition":"selected_load_bearing","contribution_type":"SAFETY_GUARDRAIL","adaptation_or_rejection":"A resident-readable report states loadings, uncertainty, couplings, prohibited interpretations, residuals, and validity limits.","counterfactual_removal":"Removing it breaks traceability and increases the chance that tentative modes become authoritative labels."},{"slug":"spectral_gap_monitor","disposition":"selected_load_bearing","contribution_type":"SAFETY_GUARDRAIL","adaptation_or_rejection":"Re-estimates gaps, mode alignment, and residuals; threshold failure automatically suspends targeting.","counterfactual_removal":"A once-valid reduced model could continue directing action after its basis loses stability."}],"causal_chain":["Coupled social conditions evolve through a locally repeatable transition that univariate alerts incompletely represent.","Decomposition identifies provisional combinations that amplify, persist, or decay and traces them to observable measures.","Sensitivity and residual tests distinguish outcome-relevant modes from loud but irrelevant or poorly reconstructed modes.","Resident-approved action bundles are chosen for their predicted leverage on a concerning mode, with coupling and subgroup constraints.","A 12-week, four-neighborhood randomized-offer pilot compares the bundle with ordinary practice using preregistered modal, raw-outcome, residual, and harm endpoints.","If the targeted mode changes before or alongside improved raw outcomes without excess harms, the lever gains bounded support; otherwise it is rejected or revised."],"baseline":"Ordinary practice uses separate trend dashboards, complaint or incident thresholds, periodic surveys, and professional judgment to deploy general outreach or enforcement after visible deterioration.","nearest_rival":"A preregistered multilevel time-series risk model using the same raw variables and interactions but no invariant-mode decomposition or modal intervention map.","authority_safety":{"affected_parties":["Residents, including nonparticipants and marginalized subgroups","Informal leaders and neighborhood organizations","Frontline outreach and municipal service staff","People represented in administrative or observational data"],"decision_authority":"Data custodians authorize lawful data use; a compensated resident oversight board controls neighborhood interpretation and approves voluntary action bundles; municipal managers control only services within their mandate. Analysts have no unilateral deployment authority.","authorized_first_step":"After retrospective holdout eligibility is met, the resident board may authorize a 12-week randomized offer of a resident-chosen, noncoercive service-and-contact bundle in four consenting neighborhoods, with aggregate measurement and an ordinary-practice comparator.","excluded_actions":["Individual risk scoring or public neighborhood labeling","Police deployment, sanctions, benefits restriction, or surveillance triggered by a mode","Covert manipulation or engineered conflict","Inference of fixed cultural traits from mode loadings","Deployment outside the validated window or to nonconsenting communities"],"halt_rollback":"Immediately suspend recommendations if the spectral gap or alignment gate fails, structured subgroup residuals exceed tolerance, privacy is breached, participation becomes coercive, or any monitored harm exceeds its preregistered boundary. Revert to ordinary practice, stop new data collection where required, and submit results to resident review."}},"negative_tests":{"strongest_counterevidence":"The same mode basis may fail to replicate across periods or neighborhoods because measurement changes, reflexive behavior, policy shocks, and contested meanings dominate; the nearest-rival model may predict equally well with less interpretive risk.","analogy_break":"Unlike a physical linear system, a community observes and reacts to classifications, indicators are institutionally produced, membership changes, causal relations are nonlinear, and approximate modes need not be independent or invariant.","failure_condition":"The proposal fails if no stable, interpretable, outcome-relevant coupled direction survives holdout reconstruction, uncertainty, subgroup residual, spectral-gap, and drift gates.","problem_falsifier":"The diagnosed problem is falsified if adverse outcomes are adequately and consistently anticipated by one directly observable variable or ordinary thresholds, with no repeatable coupled transition adding predictive or explanatory value.","intervention_falsifier":"The intervention conjecture is falsified if the modal bundle does not outperform the nearest rival on preregistered raw outcomes, fails to move the targeted mode, changes outcomes without the predicted modal mediation, or increases subgroup harms.","risks":["Reifying statistical modes as cultural essences","Neighborhood stigma and resource withdrawal","Ecological inference from aggregates to individuals","Privacy loss through small-cell data","Confounding and false causal interpretation","Mode instability after shocks or policy changes","Spillovers that contaminate comparison neighborhoods","Optimization of measured indicators rather than lived conditions"]},"null_rationale":null,"classification":{"candidate_kind":"DOMAIN_TRANSFER","prior_art_status":"UNSEARCHED","evidence_maturity":"HYPOTHESIS"},"revision_change_log":{"revision_kind":"ORIGINAL","prior_problem_id":null,"prior_causal_lever_id":null,"problem_changed":false,"causal_lever_changed":false,"conceptual_changes":[],"operational_changes":[],"repairs_addressed":[]},"confidence":0.73,"generator_notes":"Candidate fit depends on demonstrating a locally stable transition rather than merely correlated survey dimensions. All empirical performance and causal claims remain hypotheses pending the bounded comparison."}