{"schema_version":1,"research_id":"eoa_inverse_innovation_exp04_external_evaluation_20260802","source_assessment_id":"invariant_mode_decomposition_design__political_science:RETRIEVAL_FIRST:v0","cell_id":"invariant_mode_decomposition_design__political_science","search_queries":["site:oig.doc.gov OIG-22-014-I H referrals 29 percent one year","site:ombudsman.parliament.nz official information delay complaints 20 25 percent tracking systemic issues","public administration network interventions pragmatic action network science 2109.08197","key player problem Borgatti network node removal centrality","site:oig.doc.gov/about/faqs audits evaluations annual risk assessment Commerce OIG","site:oecd.org \"Improving Collaboration in the Public Sector Audit System in Brazil\"","\"The Network Organisation of Consumer Complaints\" 1008.1815","site:opm.gov salary table 2026 DCB"],"sources":[{"source_id":"S1","title":"The Department Must Improve Processes for Addressing and Managing ‘H’ Referrals","publisher":"U.S. Department of Commerce Office of Inspector General","url":"https://www.oig.doc.gov/OIGPublications/OIG-22-014-I.pdf","source_class":"GOVERNMENT_OR_REGULATOR","publication_date":"2021-12-16","accessed_at":"2026-08-02","claims_supported":["Commerce lacked effective department-wide policies, procedures, and timeliness monitoring for hotline referrals.","Approximately 29 percent of reviewed referrals had remained outstanding for more than one year despite a general 60-day response or extension requirement.","The report connects delayed referral handling to diminished effectiveness, accountability, and the risk that alleged wrongdoing is not fully addressed.","The records contained bureau, referral-age, inquiry-status, and disposition information, demonstrating that structured administrative referral data can exist."]},{"source_id":"S2","title":"Ombudsman’s approach to official information delay complaints","publisher":"Office of the Ombudsman, New Zealand","url":"https://www.ombudsman.parliament.nz/resources/ombudsmans-approach-official-information-delay-complaints","source_class":"OFFICIAL_GUIDANCE","publication_date":"2025-07-31","accessed_at":"2026-08-02","claims_supported":["Delay complaints consistently comprise approximately 20–25 percent of official-information complaints received by the office.","Formal investigation of every delay complaint had not, by itself, produced meaningful improvement after three years.","The refreshed approach explicitly emphasizes early resolution and identifying, monitoring, and resolving potential systemic issues."]},{"source_id":"S3","title":"Audits and Evaluations FAQs","publisher":"U.S. Department of Commerce Office of Inspector General","url":"https://www.oig.doc.gov/about/faqs/faqs-audits-evaluations/","source_class":"OFFICIAL_GUIDANCE","publication_date":"","accessed_at":"2026-08-02","claims_supported":["Commerce OIG conducts annual program risk assessments and uses them to develop its audit plan.","Audit selection may reflect legal mandates, requests, management priorities, prior findings, and elevated program or office risk.","OIG management approves audit and evaluation projects, and official objectives can include prospective analysis and assessments of internal control.","Audits require sufficient, appropriate evidence under Government Auditing Standards."]},{"source_id":"S4","title":"Network Interventions: Applying Network Science for Pragmatic Action in Public Administration and Policy","publisher":"arXiv","url":"https://arxiv.org/abs/2109.08197","source_class":"PRIMARY_RESEARCH","publication_date":"2021-09-16","accessed_at":"2026-08-02","claims_supported":["Network interventions purposefully use network data to identify strategies for improving performance or other outcomes.","The framework expressly extends network-intervention strategies to public-sector interorganizational and governance networks.","The paper establishes a close conceptual precedent for intervention-oriented use of public-sector network data."]},{"source_id":"S5","title":"El problema del actor clave","publisher":"Redes: Revista Hispana para el Análisis de Redes Sociales, Universitat Autònoma de Barcelona","url":"https://revistes.uab.cat/redes/article/view/v24-n2-borgatti","source_class":"PRIMARY_RESEARCH","publication_date":"2013-12-16","accessed_at":"2026-08-02","claims_supported":["The Key Player Problem selects nodes according to the intended network outcome rather than relying on generic importance.","It formalizes node selection for optimal diffusion or for maximum disruption or fragmentation after node removal.","It finds that standard off-the-shelf centrality measures are not generally optimal for these intervention objectives."]},{"source_id":"S6","title":"The network organisation of consumer complaints","publisher":"arXiv","url":"https://arxiv.org/abs/1008.1815","source_class":"PRIMARY_RESEARCH","publication_date":"2010-08-10","accessed_at":"2026-08-02","claims_supported":["The study constructs a weighted complaint network from Brazilian Department of Justice administrative records.","It relates complaint-network structure to complaint resolution and demonstrates dimensional reduction of complaint features.","It establishes prior practice for representing administrative complaint records as a network, although its edges are complaint-category-to-company relationships rather than interoffice referrals."]},{"source_id":"S7","title":"Improving collaboration in the public sector audit system in Brazil","publisher":"OECD","url":"https://www.oecd.org/en/publications/auditing-decentralised-policies-in-brazil_30023307-en/full-report/improving-collaboration-in-the-public-sector-audit-system-in-brazil_4f1436fc.html","source_class":"OFFICIAL_GUIDANCE","publication_date":"2020-11-26","accessed_at":"2026-08-02","claims_supported":["Public audit institutions already use collaborative networks, evidence-based risk assessment, and shared approaches to audit prioritization.","The OECD recommends mechanisms for collaborative audit selection, a dedicated team, shared prioritization, and preliminary assessment.","The guidance identifies information sharing, confidentiality, privacy, differing standards, and institutional capacity as implementation constraints."]},{"source_id":"S8","title":"Salary Table 2026-DCB","publisher":"U.S. Office of Personnel Management","url":"https://www.opm.gov/policy-data-oversight/pay-leave/salaries-wages/salary-tables/26Tables/html/DCB.aspx","source_class":"OFFICIAL_ORGANIZATION_DATA","publication_date":"2026-01","accessed_at":"2026-08-02","claims_supported":["The 2026 Washington-Baltimore-Arlington annual salary range is $102,415–$133,142 for GS-12, $121,785–$158,322 for GS-13, and $143,913–$187,093 for GS-14.","These rates provide a public labor-cost anchor before benefits, overhead, contracting premiums, security engineering, and governance costs."]}],"problem_evidence":{"support":"STRONG","rationale":"Official evidence shows consequential, persistent referral delay and weak monitoring in a concrete federal system: 29 percent of Commerce H referrals were older than one year, and inconsistent bureau processes impaired accountability. Independent official evidence from New Zealand shows that delay is also a recurring complaint-management problem and that case-by-case investigation alone did not yield systemic improvement. The sources establish the general problem, but they do not establish the candidate’s narrower premise that repeated office-to-office transfers form a stable dominant mode.","source_ids":["S1","S2"]},"stakeholder_evidence":{"support":"MODERATE","rationale":"Commerce OIG is an identifiable authorizer with audit-planning authority, an established annual risk-assessment workflow, permission to conduct prospective analysis, and a documented need for better referral monitoring. OECD guidance shows institutional demand for evidence-based, collaborative audit prioritization. No source expresses demand for spectral referral analysis specifically, documents a committed adopter, or commits data, staff, or funding.","source_ids":["S1","S3","S7"]},"prior_art":{"proximity":"SUBSTANTIAL_COLLISION","closest_analogues":[{"name":"Key Player Problem","similarity":"Already replaces generic centrality with outcome-dependent counterfactual node selection, closely anticipating the candidate’s perturb-and-rank logic.","remaining_difference":"Its disclosed objectives concern diffusion or network fragmentation, not marginal reduction of a longitudinal responsibility-transfer eigenmode validated against future unresolved referrals.","source_ids":["S5"]},{"name":"Public-sector network-intervention framework","similarity":"Already proposes purposeful use of governance-network data to select actors or structural interventions that improve public-sector outcomes.","remaining_difference":"It does not specify complaint-referral edges, a leading responsibility-transfer mode, audit targeting, or prospective unresolved-case validation.","source_ids":["S4"]},{"name":"Administrative complaint-network analysis","similarity":"Already constructs weighted networks from government complaint records and relates network structure to resolution while using dimensional reduction.","remaining_difference":"Its network links complaint categories to companies; it does not model interoffice responsibility transfers or simulate feasible changes to closure and onward-transfer behavior.","source_ids":["S6"]},{"name":"Risk-based and collaborative public-sector audit selection","similarity":"Official practice already combines risk assessment, institutional information, preliminary analysis, and coordinated prioritization to select audit work.","remaining_difference":"The opened guidance does not use a referral-network eigenmode, node-level modal sensitivity, or a prospective comparison with complaint volume, degree, and existing risk scores.","source_ids":["S3","S7"]}],"distinctive_claim_remaining":"In longitudinal agency referral networks, ranking audit targets by each node's marginal reduction of the leading responsibility-transfer mode predicts future unresolved or repeatedly transferred cases better than complaint volume, degree centrality, generic eigenvector centrality, and existing risk scores.","confidence":"HIGH"},"implementation_evidence":{"support":"MODERATE","rationale":"The analysis is computationally plausible because weighted administrative complaint networks, public-sector network interventions, and counterfactual node-selection methods are established. A shadow study fits an existing prospective-analysis and risk-assessment workflow. However, the decisive data and validity conditions remain unverified: S1 demonstrates referrals from OIG to bureaus, not a complete longitudinal chain of repeated bureau-to-bureau transfers; directed networks may be reducible or have a small spectral gap; recording changes can create spurious modes; and simulated edge perturbations are predictive scenarios rather than identified causal effects. Access control, privacy review, data authority, human review, residual checks, mode-drift suspension, and a prohibition on treating rank as evidence of wrongdoing are required.","source_ids":["S1","S3","S4","S5","S6","S7"]},"scores":{"meaningful_impact":{"score":4,"rationale":"Delayed referrals can impair effective inquiry and accountability, and two official systems describe persistent timeliness or systemic-resolution problems. Realized impact from changing audit prioritization remains untested.","source_ids":["S1","S2"]},"stakeholder_pull":{"score":3,"rationale":"An identifiable OIG already performs risk-based audit planning and prospective analysis, while OECD guidance calls for evidence-based collaborative prioritization. There is no expressed request or commitment for this particular method.","source_ids":["S3","S7"]},"incremental_advantage":{"score":2,"rationale":"Outcome-specific modal sensitivity could outperform workload and generic centrality, but Key Player methods already embody outcome-specific perturbation and no empirical incremental performance has been shown.","source_ids":["S4","S5"]},"distinctiveness_plausibility":{"score":2,"rationale":"The narrow integration was not found in the bounded search, but nearly every component has close precedent in complaint-network analysis, network intervention, key-player selection, and risk-based audit planning.","source_ids":["S3","S4","S5","S6","S7"]},"technical_implementability":{"score":3,"rationale":"Sparse weighted-network estimation and perturbation are routine research operations, but data completeness, reducibility, spectral separation, temporal drift, and realistic perturbation definitions are material unresolved conditions.","source_ids":["S1","S4","S5","S6"]},"adoption_authority_feasibility":{"score":3,"rationale":"OIG management has recognizable authority over project approval and audit selection, so an advisory shadow analysis has a plausible authorization path. Data-use and privacy approvals have not been verified.","source_ids":["S3","S7"]},"evidence_readiness":{"score":2,"rationale":"Official records demonstrate potentially useful structured fields, but no authorized candidate dataset, transfer-chain prevalence estimate, spectral diagnostics, or temporal-holdout result is available.","source_ids":["S1"]},"safety_net_benefit":{"score":3,"rationale":"A shadow-only study with human authority, residual and drift stops, and no misconduct inference could reduce the risk of overlooking structural bottlenecks while preserving existing planning if the model fails. Ranking could still stigmatize lawful specialist or high-workload offices.","source_ids":["S1","S3"]},"scalability":{"score":2,"rationale":"The computation should scale, but cross-system diffusion is constrained by heterogeneous referral definitions, legal mandates, access rules, coding practices, and institutional capacity.","source_ids":["S4","S7"]}},"score_confidence":"MODERATE","costs":{"first_evidence":{"band_2026_usd":"50K_TO_250K","scope":"Four-to-six-month, single-system shadow feasibility study: data inventory, approvals, preregistration, network construction, baseline models, temporal holdout, stability and residual analysis, and an internal report.","confidence":"MODERATE","assumptions":["Approximately 0.5–1.25 FTE of GS-12-to-GS-14-equivalent analyst, data-steward, privacy, and oversight time.","Approved structured fields already exist and do not require large-scale narrative extraction or record digitization.","Secure analytic infrastructure is already available.","The estimate is resource-equivalent and excludes the cost of the underlying case-management system."],"source_ids":["S1","S8"]},"initial_deployment_startup":{"band_2026_usd":"250K_TO_1M","scope":"Build and validate a repeatable access-controlled pipeline, data-quality controls, versioned operator specifications, audit logs, review interface, documentation, threat/privacy assessment, and governance procedures for one oversight organization.","confidence":"LOW","assumptions":["Approximately 2–4 FTE-equivalent staff for 6–12 months across analytics, engineering, data stewardship, security/privacy, legal review, and audit operations.","No procurement of a new enterprise case-management platform is required.","The system remains advisory and does not automatically open or prioritize audits."],"source_ids":["S3","S7","S8"]},"operational_launch":{"band_2026_usd":"250K_TO_1M","scope":"One-system controlled launch after successful evidence: parallel-run validation, user training, independent model review, incident and rollback procedures, accessibility, and integration into an existing audit-planning cycle.","confidence":"LOW","assumptions":["Launch covers one oversight system and one planning workflow.","At least two reporting cycles are run in shadow or parallel before the ranking informs human deliberation.","Security authorization and data agreements can be completed without major platform replacement."],"source_ids":["S3","S7","S8"]},"annual_recurring":{"band_2026_usd":"250K_TO_1M","scope":"Ongoing data refresh, quality review, spectral-gap and drift monitoring, temporal revalidation, access control, security maintenance, model documentation, appeals or correction handling, and annual independent review.","confidence":"LOW","assumptions":["Approximately 1.5–3.5 FTE-equivalent recurring effort across analytics, engineering, data stewardship, privacy/security, and audit governance.","Compute is modest relative to labor for a sparse organizational network.","Expansion to multiple legally distinct oversight systems would require separate data mappings and approvals and could exceed this band."],"source_ids":["S7","S8"]}},"verified_pipeline_gates":{"externally_supported_problem":{"status":"YES","reason":"Official, direct evidence documents material referral delays, inconsistent processes, missing monitoring, and recurring systemic complaint-delay concerns.","source_ids":["S1","S2"]},"externally_credible_adopter_or_authorizer":{"status":"YES","reason":"Commerce OIG is an identifiable authority that approves audit projects, conducts annual risk assessment, and may undertake prospective analysis; analogous audit institutions are encouraged to use evidence-based collaborative prioritization.","source_ids":["S3","S7"]},"distinct_testable_incremental_claim":{"status":"YES","reason":"The remaining claim specifies a ranking method, four comparators, future unresolved or repeated-transfer outcomes, and a temporal holdout on which incremental predictive value can be falsified.","source_ids":["S3","S5"]},"bounded_next_evidence_step":{"status":"YES","reason":"A single-system, structured-field-only, non-deployment temporal-holdout study can be preregistered and completed without changing audit selection or referral routing.","source_ids":["S1","S3"]},"no_unresolved_safety_or_authority_stop":{"status":"UNCERTAIN","reason":"OIG has plausible project authority and the study can remain advisory, but written data-use authority, privacy/security approval, rules for office-level ranking disclosure, and confirmation that referral records may be linked longitudinally have not been obtained.","source_ids":["S3","S7"]},"credible_cost_scope_and_range":{"status":"YES","reason":"The four estimates state their organizational scope and FTE assumptions and are anchored to official 2026 federal salary data, although integration and authorization costs remain low-confidence.","source_ids":["S8"]}},"next_evidence_step":"Obtain written participation and data-use approval from one oversight system, then run a six-month, non-deployment study on at least 24 consecutive monthly reporting periods using approved structured fields only. Before access to the final six months, preregister nodes, directed edges, closure handling, repeat-transfer definition, a policy-plausible perturbation no larger than 10 percent of an office's eligible onward-transfer or overdue-case rate, spectral-gap and mode-drift thresholds, missing-data exclusions, residual tolerance, the audit-list size k, and the primary endpoint. Fit on months 1–18 and evaluate on untouched months 19–24. Compare the modal-sensitivity ranking with complaint volume, directed and weighted degree, generic eigenvector centrality without outcome-specific perturbation, and the existing risk score. Use future unresolved-at-60-days and repeatedly transferred cases as outcomes; use NDCG@k as the primary ranking metric and report top-k capture, calibration, block-bootstrap intervals, subgroup-by-routing-function diagnostics, rank stability, residual structure, and sensitivity to reasonable edge definitions. Progress only if the method achieves at least a 10 percent relative NDCG@k improvement over the strongest comparator with a block-bootstrap 95 percent confidence interval for the absolute difference above zero, rank correlation of at least 0.70 across reasonable specifications, and all preregistered gap, drift, residual, data-quality, privacy, and human-review checks. Falsify the incremental claim if performance does not clear that bar; reject modal prioritization if no reproducible coupled transfer structure exists, the leading mode is unstable, residuals remain consequentially structured, or realistic perturbations do not preserve the ranking. Do not expose identifiable rankings outside the authorized review group or use results for live audit selection.","blocking_evidence":["No written adopter commitment, data-use authorization, privacy/security approval, or funding commitment has been observed.","The opened official evidence documents OIG-to-bureau referrals but does not establish that repeated office-to-office transfers are frequent, consistently coded, or longitudinally linkable.","No candidate dataset has demonstrated a reproducible leading mode, adequate spectral separation, stable mode identity, or acceptable held-out reconstruction residuals.","No temporal-holdout evidence shows incremental predictive value over complaint volume, degree, generic eigenvector centrality, or the existing risk score.","No evidence establishes that the simulated closure or onward-transfer perturbations correspond to feasible organizational changes or identify causal effects.","The risk that lawful specialization, mandatory routing, or differential recordkeeping produces stigmatizing office rankings has not been empirically assessed.","Transportability across oversight systems with different mandates, schemas, privacy rules, and referral practices is unknown."],"research_disposition":"PARTNERED_RESEARCH_PROGRAM","world_novelty_boundary":"This bounded ordinary-web review through 2026-08-02 found established administrative complaint-network analysis, outcome-specific key-player perturbation, public-sector network-intervention frameworks, risk-based audit planning, collaborative audit selection, and official referral tracking. It did not find an opened source implementing the complete combination of a longitudinal interoffice referral operator, leading responsibility-transfer mode, policy-plausible node sensitivity, and prospective audit-target validation against future unresolved or repeated transfers. Any distinctiveness is therefore limited to that narrow integration and its unproven incremental-performance claim. World novelty, patentability, freedom to operate, market size, and realized impact were not measured.","arm":"RETRIEVAL_FIRST","candidate_version":0,"controller_recommendation":{"action":"STOP_EMPIRICAL_RESEARCH_NEEDED","repairable":true,"material_progress_observed":true,"progress_targets":["Secure a named oversight partner, written data-use authority, privacy/security approval, and a documented human-review and nondisclosure workflow.","Measure the prevalence, completeness, and longitudinal linkability of repeated interoffice transfers before fitting a modal model.","Preregister operator construction, realistic perturbations, spectral-gap, drift, residual, missingness, performance, fairness, and rollback thresholds.","Run the specified untouched temporal holdout against complaint volume, directed and weighted degree, generic eigenvector centrality, and the existing risk score.","Demonstrate that results are not artifacts of lawful specialization, mandatory routing, reporting-period choice, missing data, or reasonable alternative edge definitions.","Obtain independent review confirming that rankings remain advisory, are not evidence of wrongdoing, and can be withdrawn without changing established audit authority."],"reason":"Bounded web research has established a consequential problem, a credible authorizer, substantial prior-art collision, and a narrow falsifiable residual claim. The remaining uncertainties depend on access-controlled administrative records, institutional authorization, privacy review, and prospective or temporal testing; they cannot be resolved by further ordinary web search."}}