{"schema_version":1,"experiment_id":"eoa_inverse_innovation_exp05_complete_proposal_portfolio20_20260803","cell_id":"invariant_mode_decomposition_design__literature_literary_theory","arm":"COMPLETE_PROPOSAL_PORTFOLIO","candidate_id":"cand_critical_edition_dependency_modes_002","proposal_index":2,"version":0,"title":"Spectral Dependency Triage for Critical-Edition Notes","problem":"Before a critical edition is frozen, a correction to one textual claim can create rechecking obligations across chains of glosses, chronology notes, source attributions, cross-references, and interpretive commentary. Editors can inspect direct backlinks or conspicuous notes yet miss combinations of claims through which verification deficits persist or recirculate. The relevant unit is therefore not an individual note but a dependency mode distributed across many notes.","actors":["General editor, who controls the edition and approves the review scope","Section editors and textual scholars, who own and adjudicate individual notes","Dependency mapper, who records explicit claim-to-claim reliance without judging literary merit","Copyeditors or research assistants, who conduct authorized source checks","Independent adjudicator, who determines whether a propagated rechecking obligation represents a genuine inconsistency"],"observable_state":"The edition is represented as a directed claim-dependency graph. Each node is an independently checkable claim appearing in a note, gloss, chronology entry, apparatus entry, or cross-reference. A directed edge from claim j to claim i records that i would require rechecking if j changed; its weight records a preregistered dependency class such as direct entailment, date inheritance, identity resolution, quotation attribution, or interpretive reliance. The state vector assigns each node a verification-deficit value. A column-normalized propagation matrix maps one round of unresolved deficits into the next round of dependent rechecking obligations. Observed correction cascades and adjudicated conflicts provide raw behavior against which modal predictions are tested.","consequence":"A source correction can be resolved locally while leaving downstream notes mutually inconsistent, or limited review capacity can be spent on highly visible claims that do not control the persistent dependency directions. The edition may then reach freeze with unresolved contradictions between its apparatus, chronology, and commentary.","affected_objective":"Prioritize bounded pre-freeze verification so that propagated claim dependencies are reconciled while preserving traceability, scholarly judgment, and coverage of consequential behavior outside the retained modes.","intervention":"Create a claim-level dependency ledger for a bounded section of the edition and assemble its rechecking-propagation operator. Decompose the operator, classify persistent or weakly damped dependency modes, and identify the claims with the largest loadings in action-relevant modes. Simulate verification or correction of candidate claims one at a time to determine which checks reduce adjudicated downstream conflict rather than merely lowering a centrality score. The general editor then authorizes a finite verification batch organized around the selected mode, while a residual queue preserves omitted claims. Recompute the operator after accepted corrections or structural note changes; suspend modal triage if the basis, spectral separation, or reconstruction performance ceases to satisfy the declared thresholds.","structural_mapping":[{"archetype_element":"Transformation Scope","domain_realization":"The transformation is the propagation of a verification deficit through explicit claim dependencies during one rechecking round in a bounded section of a critical edition."},{"archetype_element":"State-Vector Definition","domain_realization":"Each coordinate represents the unresolved verification deficit of one independently checkable editorial claim."},{"archetype_element":"Invariant Mode Basis","domain_realization":"Eigenvectors of the propagation operator represent distributed patterns of rechecking obligation that approximately retain their direction across repeated dependency traversal."},{"archetype_element":"Modal Gain Spectrum","domain_realization":"Eigenvalues characterize how quickly each dependency pattern decays, persists, oscillates between linked claim groups, or grows under the fitted propagation rule."},{"archetype_element":"Dominant Mode Selection Rule","domain_realization":"A mode becomes actionable only if it is stable under plausible edge-weight perturbations, predicts held-out correction cascades better than direct-link rivals, and contains claims whose adjudicated consequences meet the editor's review threshold."},{"archetype_element":"Stable/Unstable Mode Partition","domain_realization":"Rapidly decaying modes are assigned to ordinary local checking; persistent, weakly damped, or growing modes receive explicit cross-note review."},{"archetype_element":"Modal Intervention Map","domain_realization":"Candidate source checks and claim corrections are projected into modal coordinates to estimate which verification actions alter each persistent dependency pattern."},{"archetype_element":"Reconstruction Residual Check","domain_realization":"Observed rechecking cascades are reconstructed from retained modes, while omitted and poorly reconstructed claims remain visible in a separate residual queue."},{"archetype_element":"Mode-Coupling Register","domain_realization":"Near-degenerate modes, shared foundational sources, and cross-section dependencies are registered so apparently separate review clusters are not treated as independent."},{"archetype_element":"Spectral Gap Threshold","domain_realization":"The retained-mode cutoff is used only when the separation between the final retained mode and first omitted mode is stable under resampling and edge-weight perturbation."},{"archetype_element":"Mode Drift Monitor","domain_realization":"The spectrum, modal directions, and claim loadings are recalculated after accepted corrections, rewritten notes, or changes to the dependency ledger."},{"archetype_element":"Interpretation Scope Contract","domain_realization":"Modal loadings indicate structural rechecking leverage within the mapped section; they do not measure a note's literary quality, interpretive truth, canonical importance, or value to readers."}],"mechanism_mapping":[{"mechanism_slug":"eigendecomposition_workflow","role":"Factor the explicit claim-dependency propagation matrix into its complete invariant-mode basis and gain spectrum.","counterfactual_removal":"Without the full decomposition, the team cannot distinguish multiple persistent dependency patterns or inspect behavior omitted by a dominant-only ranking."},{"mechanism_slug":"modal_stability_analysis","role":"Classify dependency modes by their behavior under repeated propagation and attach the classification to the current ledger and weighting regime.","counterfactual_removal":"Without stability classification, a large current mode could be confused with a persistent rechecking pattern, while a smaller weakly damped mode could be overlooked."},{"mechanism_slug":"network_spectral_centrality_analysis","role":"Translate the loadings of action-relevant network modes into a claim-level verification map.","counterfactual_removal":"Without this mapping, the distributed mode would not identify which concrete claims can be checked or adjudicated."},{"mechanism_slug":"modal_sensitivity_sweep","role":"Simulate verifying or correcting each candidate claim and measure changes in predicted downstream conflicts, review effort, and other modes.","counterfactual_removal":"Without outcome-based perturbation, the largest-loading claim would be treated as the best intervention target even when checking it has little practical consequence."},{"mechanism_slug":"residual_reconstruction_test","role":"Compare retained-mode reconstructions with observed or seeded correction cascades and preserve structured omissions in the review queue.","counterfactual_removal":"Without this test, modal compression could hide a low-amplitude dependency chain containing a consequential inconsistency."},{"mechanism_slug":"spectral_gap_monitor","role":"Track whether retained modes remain separated and directionally stable as the apparatus changes.","counterfactual_removal":"Without ongoing gap and drift checks, the team could continue using a verification order derived from an obsolete dependency structure."},{"mechanism_slug":"spectral_decomposition_report","role":"Document each selected mode, its claim loadings, sensitivity results, couplings, residuals, and prohibited interpretations for editorial decision-makers.","counterfactual_removal":"Without a bounded report, structural leverage scores could be mistaken for rankings of scholarly importance or evidentiary truth."}],"causal_chain":["Claims in a critical edition depend on other claims through attribution, chronology, identity, quotation, cross-reference, and interpretation links.","A correction creates rechecking obligations that traverse those links over successive dependency rounds.","Some combinations of claims preserve or amplify verification deficits more strongly than isolated notes or direct-link counts reveal.","Decomposition of the propagation operator exposes those distributed dependency modes and their scalar persistence.","Stability, spectral-gap, and reconstruction tests determine whether a reduced modal account is usable for the bounded section.","Claim-level sensitivity simulations distinguish structurally prominent claims from checks that actually alter adjudicated downstream conflicts.","The general editor authorizes a finite verification batch targeting the supported mode while retaining a residual queue for behavior outside it.","Accepted corrections change the dependency structure, triggering re-estimation before the modal ordering can be reused."],"baseline":"The baseline is a flat pre-freeze proofing queue supplemented by manual backlink searches. Editors prioritize conspicuous notes, recently changed passages, and direct references, then resolve discovered inconsistencies through ordinary scholarly judgment.","nearest_rivals":["Direct backlink traversal, which follows every immediate dependent of a changed claim but does not rank persistent combinations produced by repeated propagation.","Transitive-closure or rule-based dependency checking, which can enumerate all reachable claims but does not distinguish strongly persistent paths from weak or immaterial ones under a fixed review budget.","Degree centrality, which prioritizes claims with many direct connections but ignores whether those connections lead into other structurally influential claims.","A static PageRank-style claim ranking, which can identify network prominence but does not by itself connect modal gains to observed correction cascades, adjudicated conflict outcomes, residual reconstruction, or intervention sensitivity.","Ordinary section-by-section scholarly review, which preserves contextual interpretation but organizes work by editorial boundaries rather than cross-section dependency directions."],"remaining_contrastive_claim":"The proposal retains a contrastive advantage only if a stable distributed dependency mode predicts held-out rechecking cascades or adjudicated conflicts better than direct backlinks, reachability, degree, and static prominence, and if mode-based sensitivity identifies a more informative verification order under the same review limit. Otherwise the simpler rival should be used.","authority_safety":{"decision_authority":"The general editor authorizes review batches and accepts or rejects corrections; designated textual scholars adjudicate claims in their expertise. The model, dependency mapper, and assistants have no authority to determine textual truth or literary value.","authorized_first_step":"Perform a read-only pilot on one completed section using existing note text, authorized sources, and version history; compare simulated verification queues without changing the edition.","excluded_actions":["Automatically rewriting, deleting, accepting, or suppressing any note or claim","Treating modal loading or network position as a measure of scholarly quality, canonical importance, or interpretive truth","Dropping low-loading claims from ordinary proofing or the residual queue","Following dependencies into manuscripts, correspondence, or licensed sources without existing authorization","Publishing the dependency ledger or exposing confidential editorial disagreements","Reusing the fitted operator after a section rewrite or threshold failure without re-estimation"],"halt_rollback":"Halt modal triage if dependency coding agreement fails its preregistered threshold, relevant claims cannot be represented as a repeatable propagation process, spectral separation is unstable, held-out reconstruction fails, sensitivity rankings reverse under modest edge-weight changes, or the general editor withdraws authorization. Revert to the original flat review queue; because the pilot is read-only, no textual rollback is required."},"negative_tests":{"strongest_counterevidence":"The dependency graph is effectively acyclic and shallow, and exact backlink or transitive-closure review identifies all adjudicated consequences with equal or better prioritization; repeated modal propagation then adds no decision-relevant structure.","problem_falsifier":"Observed corrections produce only isolated or single-hop rechecking obligations, or no coupled direction remains stable across dependency coders, plausible weight schemes, and held-out portions of the section.","intervention_falsifier":"Under an identical verification limit, the modal queue does not identify adjudicated downstream inconsistencies earlier or more completely than direct-backlink, reachability, degree, and static-centrality queues on held-out correction cases.","risks":["Dependency weights may encode the mapper's interpretation rather than actual scholarly reliance.","Circular notes may reflect legitimate reciprocal interpretation rather than an unresolved verification loop.","A prominent mode may localize around a densely cross-referenced apparatus convention with little consequence for textual accuracy.","Low-loading claims can still carry severe attribution, rights, or factual consequences.","Modal terminology may cause editors to defer to a ranking they cannot independently inspect.","Source corrections can change the graph enough to invalidate the local operator during the review itself.","Confidential disagreements or unpublished source relationships could be exposed through the dependency ledger.","Optimizing review order may narrow attention to measurable consistency while missing prose clarity or interpretive nuance."]},"next_evidence_step":"Select one frozen chapter containing no more than 100 independently checkable claims. Have two mappers independently code claim dependencies and reconcile only according to a preregistered edge taxonomy. Use a bounded set of previously adjudicated corrections from version history, or editor-approved synthetic source changes if history is insufficient, as offline propagation cases. Fit the operator on one subset and compare modal, backlink, reachability, degree, static-centrality, and flat queues on held-out cases under the same verification limit. Require blinded scholarly adjudication of whether each predicted recheck is genuine, and make no changes to the edition during the pilot.","prior_art_status":"UNSEARCHED","diversity_from_prior_proposals":"This proposal is materially independent of proposal 1, Modal Ambiguity Checkpoint for Unreliable-Narrator Revision. Proposal 1 analyzes sequential changes in prose features to preserve an author-defined ambiguity envelope and intervenes by offering alternative edit batches. This proposal analyzes propagation through a claim-dependency network to reconcile a critical edition's scholarly apparatus and intervenes by ordering source verification and cross-note adjudication. Its actors, governed artifact, state coordinates, transformation, objective, intervention, and causal path are different: editorial-feature drift in a creative manuscript is replaced by recurrent verification obligations among scholarly claims. The present opportunity can be adopted by a critical-edition team without adopting, operating, or extending the manuscript-revision checkpoint.","revision_record":{"parent_version":null,"progress_targets_addressed":[],"conceptual_changes":[],"operational_changes":[],"evidence_changes":[],"claim_changes":[]}}