{"schema_version":1,"assessment_id":"eoa_inverse_innovation_exp03_opportunity320_20260801","source_experiment_id":"eoa_inverse_innovation_exp03_full320_20260801","cell_id":"invariant_mode_decomposition_design__criminology_forensic","archetype_slug":"invariant_mode_decomposition_design","domain_slug":"criminology_forensic","title":"Shadow detection of coupled forensic-toxicology QC drift modes","opportunity_summary":"The candidate proposes estimating locally stable combinations of run-to-run quality-control changes that may precede failures while individual indicators remain within limits. A shadow-only evaluation would compare these modal warnings with existing univariate QC and a PCA-based rival without allowing alerts to alter case results or validated acceptance criteria. Predictive existence, stability, controllability, prevalence, and distinctiveness remain unverified hypotheses.","adopter_authorizer":"A forensic toxicology laboratory is the adopter; its laboratory director and quality manager can authorize the retrospective shadow pilot, while validated-procedure governance and applicable legal review would control any operational adoption.","scores":{"meaningful_impact":{"score":4,"rationale":"Earlier recognition of unreliable positive or negative toxicology findings could materially benefit analyzed persons, investigations, courts, and laboratory operations. The consequence is serious, but the frequency and preventable share of such events are unsupported."},"stakeholder_pull":{"score":3,"rationale":"Laboratory directors and quality managers have an identifiable interest in reliable, timely, and traceable QC decisions, and the proposal addresses reruns, failures, and amended findings. No sealed evidence establishes demand, dissatisfaction with current QC, or willingness to adopt additional alerts."},"incremental_advantage":{"score":3,"rationale":"The transition-based modes could detect persistent coupled degradation that independent limits miss and offer a stability interpretation absent from the stated PCA rival. No out-of-sample evidence shows better warning accuracy, timeliness, or actionability than either comparator."},"distinctiveness_plausibility":{"score":3,"rationale":"Explicit run-transition dynamics, spectral and residual gates, and an intervention map form a testable distinction from the stated high-variance PCA rival. Prior art is unsearched, so uniqueness relative to existing multivariate laboratory QC methods is unverified."},"technical_implementability":{"score":3,"rationale":"The proposal uses routinely described calibrator, blank, control, retention-time, ion-ratio, sensitivity, rerun, and maintenance records and can operate in shadow mode. Stable transition and modal estimation from 30 training runs may be underdetermined or fragile, especially across lots, analysts, maintenance, and method regimes."},"adoption_authority_feasibility":{"score":4,"rationale":"The laboratory director and quality manager are explicitly empowered to authorize the shadow pilot, and ordinary QC remains authoritative. Later operational adoption would still require procedure validation and potentially legal or accreditation review not resolved in the packet."},"evidence_readiness":{"score":3,"rationale":"The candidate specifies a 30-run fit, 10-run holdout, comparators, outcomes, preregistered thresholds, falsifiers, and rollback conditions. Label quality, event frequency, leakage control, statistical power, and sufficiency of the small sample remain unresolved."},"safety_net_benefit":{"score":5,"rationale":"The pilot is retrospective and shadow-only, prohibits case-result or misconduct conclusions from modal alerts, preserves existing QC actions, and requires discarding the pilot basis after instability, missed failures, excessive residuals, or excessive false alerts."},"scalability":{"score":2,"rationale":"The model is deliberately limited to one instrument-method regime, while instrument, method, lot, maintenance, analyst-response, and feedback changes can rotate or invalidate modes. Scaling would require regime-specific validation, monitoring, and governance rather than straightforward replication."}},"score_confidence":"MODERATE","costs":{"first_evidence":{"band_2026_usd":"10K_TO_50K","scope":"One laboratory, one instrument-method regime, retrospective preparation of 30 runs and shadow evaluation of the next 10, including preregistration, data extraction, modeling, comparator analysis, QA review, and a decision report.","confidence":"LOW","assumptions":["Relevant QC, rerun, maintenance, and amendment records are already accessible and linkable.","No new instrument hardware or production-system integration is required.","Laboratory analysts and an independent quantitative reviewer contribute limited part-time effort.","The small study is treated as an initial falsification exercise rather than validation for operational use."]},"initial_deployment_startup":{"band_2026_usd":"50K_TO_250K","scope":"Development of a controlled shadow-monitoring workflow for one laboratory and a limited set of instrument-method regimes, including data interfaces, audit logging, model monitoring, SOPs, validation planning, training, security, and compliance review.","confidence":"LOW","assumptions":["Existing laboratory information and instrument systems can export the required run-level measurements.","Alerts remain advisory and cannot change case results or existing QC rules.","Several regimes require separate configuration and validation.","No major replacement of laboratory information systems is needed."]},"operational_launch":{"band_2026_usd":"250K_TO_1M","scope":"Validated operational launch across a medium forensic toxicology laboratory, including integration, regime-specific qualification, change control, analyst training, documentation, legal and quality review, alert-response procedures, and prospective comparative evaluation.","confidence":"LOW","assumptions":["Launch covers multiple instruments and methods within one laboratory rather than a national network.","Operational use proceeds only after predictive value and response safety are demonstrated.","Substantial partner coordination and formal validation are required because findings affect legal proceedings.","Existing QC remains available as the authoritative fallback."]},"annual_recurring":{"band_2026_usd":"50K_TO_250K","scope":"Annual operation for one laboratory, including data-pipeline support, drift and stability surveillance, regime revalidation, audit review, false-alert investigation, software maintenance, training refreshers, and outcome evaluation.","confidence":"LOW","assumptions":["The system serves several instrument-method regimes in one laboratory.","Model reviews are triggered by lots, maintenance, method changes, or detected instability.","Recurring costs exclude major new laboratory hardware and casework expansion.","Human quality oversight remains mandatory."]}},"research_burden":"HIGH","earliest_credible_horizon":"3_TO_12_MONTHS","pipeline_gates":{"recognizable_externally_supportable_problem":{"status":"UNCERTAIN","reason":"The candidate defines measurable coupled sub-threshold drift and consequential downstream outcomes, but its claim that laboratories commonly rely on separate indicators and the prevalence of uncaught coupled degradation require external support."},"identifiable_adopter_or_authorizer":{"status":"YES","reason":"The forensic toxicology laboratory is the adopter, and the laboratory director and quality manager are expressly identified as authorities for a shadow pilot."},"distinct_testable_incremental_claim":{"status":"YES","reason":"The proposal claims that gated transition modes predict failures, reruns, maintenance events, or amended findings out of sample beyond individual indicators, ordinary covariates, and a PCA-based multivariate rival."},"bounded_next_evidence_step":{"status":"YES","reason":"A one-regime 30-run retrospective fit followed by 10 shadow runs is specified, with preregistered thresholds, existing-QC and PCA comparisons, and explicit predictive and stability falsifiers."},"no_unresolved_safety_or_authority_stop":{"status":"YES","reason":"The first step preserves existing QC authority, excludes case-result and legal-evidence uses, prohibits changes to acceptance criteria, and includes explicit stop-and-discard conditions."},"implementation_cost_scope_and_range":{"status":"UNCERTAIN","reason":"A one-laboratory shadow scope can be bounded and broad resource bands can be estimated, but data-interface complexity, record quality, validation obligations, number of regimes, and legal or accreditation requirements are not specified."}},"blocking_evidence":["No evidence establishes that reproducible coupled drift modes exist or precede adverse outcomes beyond individual QC indicators and ordinary covariates.","The proposed 30-run fitting and 10-run holdout design lacks a demonstrated sampling and power basis for stable transition estimation or sufficiently frequent outcome events.","Eigenvector stability, conditioning, spectral separation, residual adequacy, and robustness to regime changes have not been demonstrated.","Retrospective labels may be confounded by maintenance responses, batch composition, documentation practices, or information leakage.","No evidence shows that permitted maintenance responses can damp an identified hazardous mode or reduce later failures.","External demand, workflow fit, validation requirements, data-access burden, and prior art are unverified."],"next_evidence_step":"With one willing laboratory, preregister the state variables, regime boundary, stability gates, alert threshold, false-alert ceiling, and outcome definitions; fit only the first 30 consecutive runs and score the next 10 in shadow mode. Compare out-of-sample warning performance and lead time against existing univariate QC, ordinary covariates, and the stated PCA rival. Falsify progression if the transition estimate fails its conditioning, spectral-gap, rotation, or residual gates; if existing failures are missed or false alerts exceed the limit; or if coupled patterns provide no reproducible incremental prediction. Do not alter case results, acceptance criteria, or required QC actions.","research_questions":["Are 30 fitting runs and 10 shadow runs sufficient for the state dimension and expected frequency of failures, reruns, maintenance events, or amended findings?","Do coupled transition modes remain identifiable and stable under bootstrap, rolling-origin, lot, analyst, maintenance, and batch-composition perturbations?","Do gated modal scores improve out-of-sample discrimination, calibration, or warning lead time over univariate QC, ordinary covariates, and PCA-based monitoring?","Can data leakage and confounding from prior maintenance or analyst responses be excluded or bounded?","Which authorized maintenance or confirmatory actions plausibly affect each mode, and can their effects later be tested against existing response rules?","What alert burden, delay, rerun rate, and missed-failure rate would laboratory leadership regard as acceptable?","What validation, accreditation, discovery, auditability, and legal-review requirements would govern operational use?","Does external prior-art research reveal equivalent transition-based multivariate QC methods already used in forensic or adjacent laboratories?","How often must models be requalified after method, instrument, reagent-lot, software, or workflow changes?"] ,"recommendation":"PARTNERED_RESEARCH","uncertainty_constraints":["Closed-book assessment: no external prevalence, prior-art, market, cost, regulatory, or realized-impact evidence was available.","The asserted common baseline and frequency of latent coupled drift are unsupported within the sealed candidate.","All predictive, stability, causal-control, and intervention benefits are hypotheses.","Cost bands are resource-equivalent planning ranges, not quotations, and depend strongly on integration and validation scope.","The first study may be too small to support more than feasibility or early falsification.","Operational use in a forensic setting carries legal and procedural constraints not resolved by the shadow-pilot design."],"closed_book_prior_art_boundary":"Prior art is unsearched and therefore unverified. This closed-book assessment makes no claim that the transition-mode approach is novel, rare, or absent from forensic toxicology, laboratory quality control, multivariate process monitoring, or adjacent fields."}