{"schema_version":1,"experiment_id":"eoa_inverse_innovation_exp04_retrieval_first_paired20_20260802","cell_id":"invariant_mode_decomposition_design__library_information_science","arm":"RETRIEVAL_FIRST","candidate_id":"H4-C0","hypothesis_id":"H4","version":0,"title":"Spectral Routing for Interlibrary Loan","problem":"Interlibrary-loan routing based on local request volume, degree, queue depth, or turnaround can miss lenders on which fulfillment depends structurally, allowing eligible assignments to concentrate on pivotal libraries and making consortium fulfillment fragile to their congestion or temporary loss.","actors":["Consortium resource-sharing governance body","Participating lending libraries","Borrowing-library ILL staff","Requesting patrons","Routing-system operator and analyst"],"observable_state":"Within a rolling directed, weighted library-to-library request-flow network, one or more eligible lenders have high dominant-mode dependency scores and their simulated removal causes a predeclared increase in unfilled or delayed requests, even when their local volume or degree is not the largest.","consequence":"Routing continues to concentrate demand on structurally pivotal lenders, increasing maximum lender-load share and the disruption produced by a lender outage or withdrawal, with possible delays for patrons and additional exception work for staff.","affected_objective":"Reduce lender-load concentration and fulfillment disruption under lender loss while keeping median fulfillment time within a predeclared noninferiority margin.","intervention":"After existing holdings, participation, service, payment, embargo, exclusion, and other eligibility rules generate lender candidates, construct a rolling request-flow operator, estimate lender scores from its dominant mode, and replay removal of high-scoring lenders against fulfillment outcomes. When a lender's replayed disruption exceeds a predeclared sensitivity threshold and the modal estimate passes gap and stability checks, diversify ordering only among already eligible lenders toward lower-dependency alternatives, subject to a predeclared median-fulfillment-time constraint. Treat the scores as routing-risk indicators rather than causal judgments about a library.","structural_mapping":[{"archetype_element":"Transformation Scope","domain_realization":"A fixed rolling window of directed, weighted request flows among consortium libraries from eligible-lender generation through assignment and recorded fulfillment outcome."},{"archetype_element":"State-Vector Definition","domain_realization":"One coordinate per consortium library node, with the edge orientation, weighting, normalization, missing-event treatment, and time window declared before analysis."},{"archetype_element":"Invariant Mode Basis","domain_realization":"The dominant eigenvector of the declared request-flow operator, oriented so its node entries represent recursively reinforced lender dependency."},{"archetype_element":"Modal Gain Spectrum","domain_realization":"The leading eigenvalue and estimated separation from competing modes, used to determine whether a single dominant dependency pattern is interpretable."},{"archetype_element":"Dominant Mode Selection Rule","domain_realization":"Use dominant-mode lender scores only when convergence, spectral-gap, and window-stability criteria are met; otherwise retain the baseline router."},{"archetype_element":"Modal Intervention Map","domain_realization":"Map high lender scores that are confirmed by node-loss replay to reduced ordering priority among otherwise eligible candidates, not to eligibility removal."},{"archetype_element":"Mode-Coupling Register","domain_realization":"Record near-degenerate modes, score instability, and lenders whose simulated rerouting shifts dependency or load onto another pivotal lender."},{"archetype_element":"Reconstruction Residual Check","domain_realization":"Compare the retained spectral representation with observed request-flow structure and inspect whether important lender-specific or route-specific behavior remains unexplained."},{"archetype_element":"Mode Drift Monitor","domain_realization":"Re-estimate node scores, mode alignment, and the spectral gap across successive rolling windows before any later operational use."},{"archetype_element":"Interpretation Scope Contract","domain_realization":"Scores describe topology-dependent routing exposure within the specified data window; they do not measure collection quality, staff performance, patron value, or intrinsic institutional importance."}],"mechanism_mapping":[{"mechanism_slug":"network_spectral_centrality_analysis","role":"Represent consortium request flows as a network operator and produce the topology-aware lender dependency ranking that local volume and degree baselines may miss.","counterfactual_removal":"Without this mechanism, the proposal reduces to existing local-metric, random, or ratio-based diversification and loses its remaining topology-aware distinction."},{"mechanism_slug":"power_iteration_probe","role":"Estimate the dominant lender-dependency mode and use convergence behavior to test whether it is sufficiently separated and stable for bounded interpretation.","counterfactual_removal":"Without this mechanism or an equivalent dominant-mode estimator and gap check, no operationally reproducible dominant score exists and an unstable mixture of near-tied modes could trigger routing changes."},{"mechanism_slug":"modal_sensitivity_sweep","role":"Replay removal or reduced availability of candidate lenders, measure fulfillment-specific disruption and latency, and identify cross-effects on other lenders before a score can trigger diversification.","counterfactual_removal":"Without fulfillment-specific sensitivity replay, centrality alone would be treated as consequence-bearing evidence, so the threshold trigger and resilience claim would be unsupported."}],"causal_chain":["Local volume, degree, queue, and turnaround metrics describe individual lenders but may not expose recursively reinforced dependencies in consortium request flows.","A router using those signals can repeatedly select an eligible lender whose network position concentrates substitution difficulty.","Congestion or loss of that lender then propagates across requests that share limited alternative paths, increasing delayed or unfilled demand and shifting load to other lenders.","A dominant-mode estimate makes the coupled dependency pattern observable, provided the mode is separated and stable in the declared rolling window.","Fulfillment-specific node-loss replay tests whether a high spectral score corresponds to consequential disruption rather than merely structural prominence.","Threshold-triggered reordering distributes eligible assignments toward alternatives only where replay indicates excessive dependency.","Reduced concentration lowers maximum lender-load share and simulated loss disruption, while the fulfillment-time constraint prevents resilience gains from being purchased through material median delay."],"baseline":"Primary comparison is the consortium's recorded or faithfully replayed existing eligible-lender ordering, including any deployed queue-depth, turnaround-time, randomization, preference-cell, or ratio-based load balancing. Named metric-only comparators are request volume and degree; all comparators use the same eligibility set and request window.","nearest_rivals":["OCLC WorldShare smart lender strings: dynamically orders eligible lenders using turnaround time and current queue depth to avoid overload, but the reviewed documentation does not disclose rolling spectral dependency, fulfillment-specific node-loss replay, or a sensitivity trigger.","ReShare ratio-based supplier load balancing: orders eligible suppliers through a load-balancing ratio, but the reviewed documentation does not make routing topology-aware or test pivotal-lender loss.","DOCLINE randomized eligible-lender routing: diversifies requests within policy-defined routing cells, but the reviewed guidance does not use dominant modes, network centrality, or node-loss sensitivity.","Distributed Location of the Critical Nodes to Network Robustness based on Spectral Analysis: spectrally ranks critical nodes and validates them through removal effects in generic networks, but does not address directed weighted ILL flows, eligible-lender assignment, lender load, fulfillment outcomes, or latency constraints."],"remaining_contrastive_claim":"Within the reviewed sources, the remaining distinction is an ILL routing policy that derives lender dependency from a rolling request-flow spectrum, confirms it through fulfillment-specific node-loss replay, and diversifies eligible assignments when that sensitivity crosses a threshold while constraining median fulfillment time.","authority_safety":{"decision_authority":"The consortium's authorized resource-sharing governance body, with participating-library approval under its existing routing and data-governance rules, sets thresholds and may authorize any pilot; the model and analyst have no independent routing or lending authority.","authorized_first_step":"Conduct a retrospective, offline counterfactual replay on a bounded historical request window using pseudonymized library-node identifiers and existing eligibility records; produce comparative metrics and uncertainty checks without changing live routing.","excluded_actions":["Changing holdings, participation, payment, embargo, exclusion, service-level, licensing, or other eligibility rules","Automatically suspending, penalizing, or labeling a library because of its spectral score","Using patron identity, request content beyond what eligibility and outcome replay require, or scores for staff or institutional performance evaluation","Changing live lender order, sending requests, or contacting patrons or libraries during the first evidence step","Treating dominant-mode scores as causal, globally valid, or reliable when convergence, spectral-gap, residual, or drift checks fail"],"halt_rollback":"Halt analysis if eligibility histories cannot be reconstructed consistently, missing outcomes make disruption unidentifiable, the dominant mode lacks a stable gap, results reverse across reasonable windows, or privacy rules cannot be met. The authorized first step makes no live change; any later separately approved pilot must stop if its fulfillment-time bound or safety checks are breached and immediately revert to the pre-pilot routing policy."},"negative_tests":{"strongest_counterevidence":"Deployed ILL systems already perform dynamic queue-aware or ratio-based load balancing and randomized diversification, while neighboring network research already combines spectral critical-node ranking with removal tests; therefore the proposal cannot claim novelty for automated ordering, load balancing, diversification, spectral criticality, or node-removal analysis separately.","problem_falsifier":"The problem is falsified for the studied consortium if observed and replayed lender loss shows no meaningful fulfillment disruption beyond what local volume, degree, queue, turnaround, or existing routing rules predict, or if no persistent concentration on structurally pivotal lenders is present.","intervention_falsifier":"The intervention is falsified if, on held-out replay or a separately authorized pilot, spectral-threshold routing fails to reduce both maximum lender-load share and simulated node-loss disruption relative to the existing router, materially increases median fulfillment time beyond the predeclared margin, or offers no reliable improvement over volume, degree, random, queue-aware, or ratio-based comparators.","risks":["The dominant eigenvector may localize on a small cluster or become unstable when the spectral gap is small.","Historical logs may omit declined, expired, rerouted, or eligible-but-unselected alternatives needed for faithful replay.","A high dependency score may reflect scarce holdings or policy constraints that routing alone cannot diversify.","Diversion can transfer load to smaller lenders or create a new pivotal node.","Offline replay may underestimate staff behavior, lender acceptance changes, and operational feedback.","Institutional scores could be misread as quality or performance rankings despite the interpretation contract.","Optimizing median fulfillment time can conceal tail delays or subgroup-specific harm, which must remain visible as secondary checks."]},"next_evidence_step":"For one consenting consortium and one fixed historical window of no more than 12 months, reconstruct eligible lender sets and outcomes, declare the flow operator and thresholds before outcome comparison, fit the spectral scores on an earlier segment, and replay the existing router, volume, degree, randomized diversification, and spectral-threshold diversification on a later held-out segment. Report maximum lender-load share, fulfillment disruption under removal of each tested lender, median fulfillment time against a consortium-set noninferiority margin, tail delays, unfilled share, score stability across windows, spectral gap, and rerouting cross-loads. The step ends with an offline go/no-go report and authorizes no live routing change.","prior_art_status":"SEARCHED_BOUNDED","revision_record":{"parent_version":null,"progress_targets_addressed":[],"conceptual_changes":[],"operational_changes":[],"evidence_changes":[],"claim_changes":[]}}