{"schema_version":1,"experiment_id":"eoa_inverse_innovation_exp03_full320_20260801","cell_id":"invariant_mode_decomposition_design__material_culture_museum_studies","trajectory_id":"R","attempt_index":0,"archetype_slug":"invariant_mode_decomposition_design","domain_slug":"material_culture_museum_studies","decision":"CANDIDATE","problem_id":"legacy_catalog_classification_repropagation","causal_lever_id":"persistent_workflow_mode_disruption","proposal":{"problem":"[HYPOTHESIS] In a museum catalog, contested or obsolete classifications can recur after record-level correction because catalog templates, authority files, synchronization, search ranking, exhibition selection, and staff routines re-propagate coupled combinations of terms. The recognizable problem is repeated classification recurrence and downstream exposure despite isolated cleanup.","actors_substrate":["Cataloged objects and their metadata fields","Authority files, templates, synchronization jobs, search, and exhibition-selection systems","Catalogers, curators, collections managers, and digital staff","Source communities and people represented by the records","Researchers and visitors encountering catalog descriptions"],"observable_state":"For each workflow cycle, a coded state vector records the prevalence and downstream appearances of classifications that authorized reviewers have identified as contested or obsolete, by field and workflow touchpoint. Audit logs record corrections, copying, reversion, search exposure, and exhibition reuse.","consequence":"Recurring classifications can continue shaping discovery, interpretation, and institutional representation; repeated manual cleanup consumes staff effort without removing propagation paths.","affected_objective":"Reduce authorized classifications' recurrence and downstream reach while preserving provenance, legitimate plural description, discoverability, and accountable human interpretation.","structural_mapping":[{"archetype_element":"Coupled transformation","domain_realization":"An estimated workflow operator maps classification prevalence across metadata fields and institutional touchpoints from one catalog cycle to the next.","claim_kind":"HYPOTHESIS"},{"archetype_element":"Surface coordinates obscure operative directions","domain_realization":"Individual erroneous fields are visible, while recurring combinations spanning templates, authority files, synchronization, and reuse may be the persistent units.","claim_kind":"INFERENCE"},{"archetype_element":"Invariant or approximate modes","domain_realization":"Modes are combinations of fields and touchpoints whose relative pattern approximately persists as workflows operate.","claim_kind":"HYPOTHESIS"},{"archetype_element":"Modal gain","domain_realization":"A mode's scalar response estimates whether its prevalence decays, persists, or grows per workflow cycle.","claim_kind":"HYPOTHESIS"},{"archetype_element":"Modal intervention","domain_realization":"Coordinated changes to the touchpoints loading on a persistent mode aim to interrupt re-propagation rather than merely edit its current records.","claim_kind":"HYPOTHESIS"},{"archetype_element":"Residual and drift governance","domain_realization":"Unexplained recurrence, mode rotation, or loss of spectral separation withdraws permission to use the modal account.","claim_kind":"INFERENCE"}],"component_map":[{"component":"Transformation Scope","status":"adapted","domain_realization":"One bounded collection, specified workflow touchpoints, catalog version, and audit interval."},{"component":"State-Vector Definition","status":"adapted","domain_realization":"Authorized classification-by-touchpoint prevalence and exposure measurements per cycle."},{"component":"Invariant Mode Basis","status":"direct","domain_realization":"Eigenvectors of the estimated workflow propagation matrix."},{"component":"Modal Gain Spectrum","status":"direct","domain_realization":"Eigenvalues interpreted as cycle-to-cycle decay, persistence, or amplification within scope."},{"component":"Dominant Mode Selection Rule","status":"adapted","domain_realization":"Retain modes exceeding pre-registered recurrence, affected-record, exposure, or sensitivity thresholds."},{"component":"Stable/Unstable Mode Partition","status":"adapted","domain_realization":"Classify modes as decaying, marginal/persistent, or growing within the fitted workflow window."},{"component":"Modal Intervention Map","status":"adapted","domain_realization":"Map each retained mode to reversible changes at its highest-leverage workflow touchpoints."},{"component":"Reconstruction Residual Check","status":"direct","domain_realization":"Reconstruct held-out recurrence and exposure; retain visible field- and subgroup-specific residuals."},{"component":"Mode Drift Monitor","status":"direct","domain_realization":"Re-estimate modes after workflow, software, policy, staffing, or vocabulary changes."},{"component":"Interpretation Scope Contract","status":"direct","domain_realization":"Modes describe propagation patterns, not cultural truth, intent, blame, or rightful terminology."},{"component":"Mode-Coupling Register","status":"direct","domain_realization":"Record cross-effects among modes when a proposed workflow change perturbs more than its target."},{"component":"Local Linearization Window","status":"adapted","domain_realization":"Validity is limited to the selected collection and stable catalog/workflow regime."},{"component":"Spectral Gap Threshold","status":"direct","domain_realization":"A pre-registered retained-versus-dropped separation required for stable mode selection."}],"mechanism_dispositions":[{"slug":"eigendecomposition_workflow","disposition":"selected_load_bearing","contribution_type":"CORE_CAUSAL","adaptation_or_rejection":"Decompose the explicit estimated workflow operator to obtain the full basis and per-cycle gains; require conditioning diagnostics.","counterfactual_removal":"Without it, there is no invariant-mode claim or gain-based propagation diagnosis."},{"slug":"modal_sensitivity_sweep","disposition":"selected_load_bearing","contribution_type":"CORE_CAUSAL","adaptation_or_rejection":"Simulate small reversible changes to each modal coordinate and workflow control, recording recurrence, exposure, and cross-mode effects.","counterfactual_removal":"Mode size would be mistaken for actionable leverage, and coordinated intervention targets would be unsupported."},{"slug":"modal_stability_analysis","disposition":"selected_load_bearing","contribution_type":"CORE_CAUSAL","adaptation_or_rejection":"Interpret gains as decaying, persistent, or growing only inside the fitted workflow regime.","counterfactual_removal":"The analysis could not distinguish self-clearing patterns from those requiring intervention."},{"slug":"mode_shape_testing","disposition":"incompatible","contribution_type":"NONE","adaptation_or_rejection":"Physical excitation and sensor-response identification do not match semantic catalog workflows.","counterfactual_removal":"No change; modes are estimated from workflow observations, not physical response."},{"slug":"network_spectral_centrality_analysis","disposition":"considered_rejected","contribution_type":"NONE","adaptation_or_rejection":"Node-importance ranking answers who is central, not which coupled classification pattern persists.","counterfactual_removal":"No change; node ranking is not the causal target."},{"slug":"power_iteration_probe","disposition":"considered_rejected","contribution_type":"NONE","adaptation_or_rejection":"Dominant-only estimation would hide secondary persistent modes and is unnecessary for the bounded operator.","counterfactual_removal":"No change because full decomposition supplies the required spectrum."},{"slug":"principal_component_analysis","disposition":"considered_rejected","contribution_type":"NONE","adaptation_or_rejection":"Variance directions need not represent cycle-to-cycle propagation directions.","counterfactual_removal":"No change; covariance structure alone cannot support the proposed causal chain."},{"slug":"reduced_order_model","disposition":"considered_rejected","contribution_type":"NONE","adaptation_or_rejection":"A runnable surrogate adds complexity before the propagation model has earned validity.","counterfactual_removal":"No loss to the bounded diagnostic and pilot."},{"slug":"residual_reconstruction_test","disposition":"selected_load_bearing","contribution_type":"TEST_DESIGN","adaptation_or_rejection":"Choose retained modes using held-out recurrence reconstruction plus structured residual checks by field and affected group.","counterfactual_removal":"There would be no empirical bound on omitted propagation behavior or safe stopping rule for compression."},{"slug":"singular_value_decomposition","disposition":"selected_supporting","contribution_type":"SAFETY_GUARDRAIL","adaptation_or_rejection":"Use singular values and conditioning as a diagnostic for transient amplification or fragile eigenvectors in a non-normal workflow operator; do not relabel singular vectors as invariant modes.","counterfactual_removal":"A numerically tidy but unstable eigendecomposition could pass without warning."},{"slug":"spectral_decomposition_report","disposition":"selected_supporting","contribution_type":"SAFETY_GUARDRAIL","adaptation_or_rejection":"Document each mode, uncertainty, coupling, residual, authorized terminology, and prohibited interpretation.","counterfactual_removal":"Tentative descriptive modes could be circulated as cultural truth or blame assignments."},{"slug":"spectral_gap_monitor","disposition":"selected_load_bearing","contribution_type":"OPERATIONAL","adaptation_or_rejection":"Track retained/dropped separation and direction drift after each audit interval.","counterfactual_removal":"The museum could continue acting after the selected basis ceased to be stable."}],"causal_chain":["Workflow dependencies repeatedly copy or select classifications across touchpoints.","Those dependencies create approximately persistent combinations not visible in isolated record counts.","Decomposition identifies combinations with persistent or growing per-cycle gain.","Sensitivity testing identifies reversible touchpoint changes that reduce the authorized outcome without unacceptable cross-effects.","A coordinated pilot weakens the targeted mode, reducing held-out recurrence and downstream exposure relative to ordinary cleanup.","Residual, gap, drift, and community-review gates determine whether to continue, revise, or retire the model."],"baseline":"Ordinary practice is issue-by-issue record correction, periodic vocabulary cleanup, and rule changes prioritized by complaints or raw error counts.","nearest_rival":"A non-spectral dependency audit maps known copying, synchronization, and authority-file rules and coordinates fixes using process tracing. It wins if explicit pathways explain recurrence as well as the modal model.","authority_safety":{"affected_parties":["Source communities and people represented","Catalog and collections staff","Curators and digital-system operators","Researchers and visitors"],"decision_authority":"The museum's catalog/data-governance authority and accountable collection leadership, jointly with appropriately authorized representatives of affected communities for culturally sensitive classifications; the model has no authority to define rightful terminology.","authorized_first_step":"Run a pre-registered, read-only analysis on one collection, then seek separate joint approval for an eight-week, internal-only reversible pilot targeting one stable mode against a matched record-level-cleanup baseline.","excluded_actions":["Deleting provenance or historical terminology","Publicly changing records without required review","Inferring intent, blame, identity, or cultural truth from a mode","Automating culturally sensitive terminology decisions","Applying results beyond the named collection or validity window"],"halt_rollback":"Halt on residual-budget failure, sub-threshold spectral gap, excessive mode drift, unstable conditioning, material cross-mode harm, or objection from an authorized affected-party reviewer. Restore the metadata snapshot and prior workflow rules; preserve an auditable record of the pilot."}},"negative_tests":{"strongest_counterevidence":"Recurrence may be driven mainly by changing policy, incentives, staffing, or one upstream authority file rather than a stable coupled operator; semantic coding and sparse audit transitions may manufacture apparent modes.","analogy_break":"Museum classifications are contested human interpretations governed by rights, provenance, and institutional power, not passive physical coordinates. Mathematical persistence neither validates a category nor determines how it should be revised.","failure_condition":"No sufficiently stationary, diagonalizable or diagnostically stable workflow operator can be estimated, or held-out residuals remain structured above tolerance.","problem_falsifier":"After controlling for one-way imports and explicit rules, corrected classifications do not recur across cycles in coupled patterns; independent source errors or one directly observable field explain nearly all recurrence.","intervention_falsifier":"With a stable pre-registered modal model and verified manipulation, the mode-targeted pilot does not reduce held-out recurrence and exposure more than matched record-level cleanup, or creates unacceptable cross-mode effects.","risks":["Reifying contested classifications through numerical representation","Mistaking descriptive association for workflow causation","Obscuring small but consequential residual patterns","Mode instability under policy or software change","Unequal community authority or consultation burden","Coordinated edits propagating unintended loss of discoverability"]},"null_rationale":null,"classification":{"candidate_kind":"MECHANISM_COMPOSITION","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":0.76,"generator_notes":"Closed-book structural transfer. Fit depends on demonstrating a bounded, sufficiently stationary workflow operator; otherwise use the non-spectral dependency-audit rival."}