{"schema_version":1,"experiment_id":"eoa_inverse_innovation_exp06_four_proposal_generalization60_20260803","cell_id":"predictive_residual_processing__criminology_forensic","arm":"COMPLETE_PROPOSAL_PORTFOLIO","candidate_id":"prp_forensic_custody_transition_residual_assurance_v0","proposal_index":2,"version":0,"title":"Residual Assurance for Physical-Evidence Custody Transitions","problem":"A forensic laboratory must preserve and review a complete custody history for each physical evidence item. Routine receipts, secured-storage movements, analyst checkouts, returns, and transfers can produce repetitive records, while a missing scan, unexpected custodian, excessive dwell time, location mismatch, seal-status change, duplicated transition, or out-of-order event may require prompt reconciliation. Reviewing every routine transition with equal attention can delay recognition of integrity-relevant discrepancies, but static exception rules can miss deviations whose meaning depends on the item’s current state and authorized workflow.","actors":["Evidence intake clerk","Evidence custodian","Forensic laboratory analyst or technician","Forensic laboratory quality manager","Submitting law-enforcement representative","Independent auditor or technical reviewer","Prosecutor and defense counsel as downstream evidence users","Court as the adjudicative authority","Person whose case may be affected by the evidence"],"observable_state":"For each evidence item, the laboratory can observe the immutable custody-event ledger; expected and recorded custodian, location, seal state, timestamp, authorization, and workflow stage; missing heartbeats or scans; structured transition residuals; source and scanner reliability; predictor and workflow versions; queue delay; residual acknowledgements; full-ledger reconciliation disagreements; fallback duration; and performance on independently seeded or labeled custody discrepancies.","consequence":"An unreconciled custody discrepancy can remain buried among routine transactions, delaying preservation, documentation, or investigation of an integrity question. Conversely, presenting every unusual transition as suspected tampering can generate unnecessary quarantines and unsupported inferences about misconduct or evidentiary reliability.","affected_objective":"Allocate quality-assurance attention promptly to custody-state discrepancies while retaining a complete, reproducible ledger and avoiding automated conclusions about tampering, admissibility, guilt, or evidentiary weight.","intervention":"Place a residual-assurance layer over, but never instead of, the append-only custody ledger. For a declared evidence class and laboratory workflow, a frozen, versioned state-transition model predicts the next authorized custody state and its validity window from the last independently confirmed full state. Each observed scan or ledger event is compared with that prediction. The comparator emits structured residuals for absent, extra, duplicated, late, out-of-order, wrong-actor, wrong-location, authorization, and seal-state differences. A precision-weighted gate considers source reliability, timing uncertainty, evidentiary-integrity consequence, and channel capacity before routing residuals to a named custodian or quality reviewer. The review message includes the model version, prior confirmed state, expected transition, actual record, uncertainty, and links to reconstruct the complete ledger. Seal breaks, unexplained access, missing evidence, unauthorized locations or actors, acquisition-channel failure, legal holds, and reviewer requests bypass residual suppression and present the full ledger immediately. Heartbeats distinguish a predicted quiet period from a failed reporting channel. Scheduled and trigger-based full-state reconciliations re-anchor every item, while an independent random audit reviews complete ledgers that produced no material residual. Validated errors may inform an offline workflow-model review, but the active model cannot update itself from ongoing cases. Version mismatch, cumulative reconstruction error, structured drift, audit disagreement, or protected-bypass failure suspends residual mode and returns the affected items to full manual reconciliation.","structural_mapping":[{"archetype_element":"Prediction target and boundary","domain_realization":"The next recorded custodian, authorized location, seal state, workflow stage, and allowable transition window for a specified physical evidence class within one laboratory’s custody process."},{"archetype_element":"Generative model state","domain_realization":"A versioned finite-state workflow model produces the authorized next-state distribution from the last independently reconciled custody snapshot without using suspect demographics, case theory, or asserted guilt."},{"archetype_element":"Expected and actual behavior","domain_realization":"The expected transition is stored before the next observation window; the actual scan or ledger event is captured separately with timestamp, source device, operator identity, authorization record, and quality metadata."},{"archetype_element":"Prediction comparator","domain_realization":"A declared comparator preserves whether the discrepancy is missing, additional, duplicated, late, out of order, wrong actor, wrong location, unauthorized, or inconsistent in seal state."},{"archetype_element":"Precision weighting","domain_realization":"Residual priority combines transition type, integrity consequence, source reliability, timestamp uncertainty, corroborating observations, and the finite review capacity rather than relying on elapsed time or magnitude alone."},{"archetype_element":"Residual propagation and reconstruction","domain_realization":"Selected residuals travel to a named reviewer with the expected and observed transition, prior confirmed state, workflow version, uncertainty, and direct access to reconstruct the entire custody ledger."},{"archetype_element":"Synchronization and validity","domain_realization":"The operational recorder and reviewer use the same workflow checksum, residuals expire after a declared validity window, and scheduled full snapshots bound accumulated state divergence."},{"archetype_element":"Raw-state audit","domain_realization":"An independent reviewer examines a random baseline plus risk-stratified complete custody ledgers, including items for which the residual channel remained quiet."},{"archetype_element":"Safety and rights bypass","domain_realization":"Missing evidence, seal changes, unexplained access, unauthorized actors or locations, reporting-channel failure, legal holds, and reviewer requests always trigger full-ledger presentation and human acknowledgement."},{"archetype_element":"Bounded learning and fallback","domain_realization":"Human-classified residuals enter an offline review of the workflow model; the active case model remains fixed, reversible, and subordinate to complete manual reconciliation."}],"mechanism_mapping":[{"mechanism_slug":"innovation_residual_filter","role":"Maintains an uncertainty-tagged estimate of each item’s current custody state and computes the innovation between the predicted and recorded transition.","counterfactual_removal":"Without a running state estimate, deviations would be assessed as isolated records rather than as discrepancies relative to the item’s reconstructed custody state."},{"mechanism_slug":"event_triggered_residual_reporting","role":"Routes a structured custody discrepancy when its reliability- and consequence-weighted score crosses the declared review threshold, while heartbeats verify that silence is operational.","counterfactual_removal":"Without event-triggered reporting, the quality channel would either repeat the complete ledger or rely on unverified silence that could conceal a failed recorder."},{"mechanism_slug":"precision_weighted_error_gate","role":"Prioritizes reliable, integrity-relevant discrepancies within a bounded reviewer queue and logs every suppressed residual for later audit.","counterfactual_removal":"Without precision and consequence weighting, noisy timing deviations could crowd out smaller but better-supported authorization or seal-state discrepancies."},{"mechanism_slug":"model_version_checksum_handshake","role":"Prevents a transition residual generated under one workflow version from being interpreted against another and tags replayed residuals with their baseline.","counterfactual_removal":"Without version compatibility checks, a legitimate transition after a workflow change could be reconstructed as an irregularity, or an invalid transition could appear legitimate."},{"mechanism_slug":"periodic_full_state_resynchronization","role":"Reconciles the predicted item state with a complete ledger and physical inventory snapshot on a schedule or earlier drift trigger.","counterfactual_removal":"Without full-state re-anchoring, one missing or misordered transition could corrupt every subsequent reconstruction indefinitely."},{"mechanism_slug":"shadow_raw_channel_sampling","role":"Independently reviews randomly selected complete ledgers, including quiet cases, to measure discrepancies suppressed or never recognized by the production predictor.","counterfactual_removal":"Without raw sampling, the system would evaluate only the discrepancy classes its current model was capable of exposing."},{"mechanism_slug":"model_drift_monitoring","role":"Detects changes in residual distributions, workflow use, recorder reliability, staleness, and realized reconciliation errors that invalidate residual-only assurance.","counterfactual_removal":"Without drift monitoring, an unofficial process change or degraded scanning channel could become a sustained source of false reassurance or nuisance alerts."},{"mechanism_slug":"raw_signal_fallback_switch","role":"Restores complete-ledger review and physical reconciliation when compatibility, observability, reconstruction, audit, or protected-event conditions fail.","counterfactual_removal":"Without a tested fallback, the predictive layer could continue filtering custody information when its state estimate was no longer trustworthy."},{"mechanism_slug":"prediction_error_review","role":"Requires a provenance-tagged human determination of whether a material residual arose from a recording error, authorized exception, workflow defect, equipment failure, model boundary, or unresolved integrity issue.","counterfactual_removal":"Without structured review, residual unusualness could be treated as proof of tampering or automatically absorbed into a changed baseline."}],"causal_chain":["A versioned workflow model predicts the next authorized custody state from the item’s last independently confirmed full state.","The append-only ledger records the actual transition independently, with provenance and an explicit missing-observation state.","A comparator converts the difference into a directional custody residual rather than a generic alarm.","Reliability and integrity consequence determine whether the residual is queued, bypassed directly to full review, or retained for audit.","The assigned reviewer receives the discrepancy with the expected baseline, observed record, uncertainty, and reconstructible full custody context.","Human review classifies and reconciles the discrepancy without allowing the residual score to determine tampering, admissibility, or evidentiary weight.","Independent full-ledger samples and periodic physical-state reconciliations reveal residuals the model suppressed, missed, or reconstructed incorrectly.","Validated residual patterns may motivate a separately approved offline workflow-model revision, while the active model remains fixed for the current evaluation.","Drift, failed observability, version mismatch, audit disagreement, or cumulative error triggers complete manual reconciliation.","The intervention is retained only if it improves the joint result of discrepancy coverage, time to acknowledgement, reconstruction fidelity, and review burden relative to full-ledger review and deterministic rules."],"baseline":"Custody assurance relies on the complete ledger, periodic physical inventory, supervisor review, and deterministic validations such as required signatures, permitted locations, and maximum checkout periods. This preserves the full record but does not represent routine transitions through a synchronized predictive state, route structured innovations, audit quiet ledgers against residual reconstruction, or use prediction errors in a governed model-review loop.","nearest_rivals":["Deterministic custody validation rules: transparent and strong for known violations, but they do not express uncertainty, reconstruct state from a synchronized predictor, or expose structured deviations outside prewritten conditions.","Full-ledger supervisory review and periodic physical inventory: provides direct completeness and remains the mandatory fallback, but allocates similar attention to routine and discrepant transitions between reconciliation points.","Barcode or RFID evidence tracking: improves observation and location recording, but sensing custody events does not itself compare them with a maintained expectation or route only calibrated residuals.","Generic anomaly detection over access logs: can rank rare transactions, but rarity is not the same as a directional custody-state mismatch and does not ensure baseline reconstruction, version synchronization, or independent quiet-ledger audits.","Workflow reminders and overdue-task alerts: address scheduled actions but do not detect wrong-state, wrong-actor, duplicated, reordered, or seal-status innovations relative to a reconstructible custody model."],"remaining_contrastive_claim":"The proposal’s testable architectural contrast is a synchronized custody-state predictor whose structured innovations become the quality-review message and bounded learning input, while complete ledger reconstruction, verified silence, independent quiet-ledger audits, protected full-context bypasses, and automatic reversion to manual reconciliation prevent the predictor from defining custody integrity by itself.","authority_safety":{"decision_authority":"The forensic laboratory quality manager may authorize or terminate a retrospective evaluation and approve workflow-model versions. Evidence custodians retain authority to quarantine and reconcile items under existing policy. Examiners may request full review at any time. Courts retain authority over admissibility and evidentiary weight; the system has no adjudicative authority.","authorized_first_step":"Conduct a read-only sandbox evaluation using synthetic custody ledgers or legally reusable records from closed matters. Freeze the workflow model and thresholds before scoring, preserve the source ledgers unchanged, inject or independently label defined transition discrepancies, and compare residual assurance with existing deterministic rules and complete manual reconciliation.","excluded_actions":["Do not alter, delete, or replace the append-only custody ledger.","Do not infer tampering, misconduct, guilt, intent, admissibility, or evidentiary weight from a residual score.","Do not automatically quarantine evidence or discipline an employee.","Do not suppress a legally required custody record from disclosure or authorized independent review.","Do not train or update the active model from unresolved or ongoing-case residuals.","Do not vary expectations using protected characteristics, suspect identity, case theory, or desired investigative outcome.","Do not deploy the evaluated layer in live custody operations during the bounded first evidence step."],"halt_rollback":"Halt residual processing for the affected item class after any protected-bypass failure, unsupported workflow version, missing heartbeat, ledger gap, unexplained reconstruction discrepancy, audit-discovered material miss, queue saturation, or loss of access to the full ledger. Restore complete manual ledger review and physical reconciliation, preserve all residual and model-version logs, invalidate the affected predictor version, and require quality-manager approval before another retrospective test."},"negative_tests":{"strongest_counterevidence":"A blinded sandbox comparison shows that deterministic rules or full-ledger review find discrepancies as quickly with equal or lower reviewer burden; residual presentation causes unsupported tampering inferences; quiet-ledger audits reveal material suppressed discrepancies; model and recorder maintenance consume the saved review capacity; or protected bypasses and reconstruction fail under realistic missing, delayed, duplicated, or reordered observations.","problem_falsifier":"The problem is falsified if routine custody records do not materially constrain quality-review attention, existing rules and reconciliation already surface integrity-relevant discrepancies within the required time, or the transition process is too irregular to support defensible state prediction.","intervention_falsifier":"The intervention is falsified if a frozen, predeclared test cannot reconstruct full custody state within the laboratory’s required fidelity, cannot maintain zero unexplained misses for protected discrepancy classes, cannot distinguish channel failure from predicted silence, or fails to improve the joint criteria of acknowledgement latency, discrepancy coverage, false escalation, and total assurance workload against the strongest rival.","risks":["A shared wrong workflow model could normalize an unauthorized but repeated practice.","Missing observations could be mistaken for valid predicted silence if heartbeats fail.","A rare but legitimate transition could be overinterpreted as suspected tampering.","Thresholds could be adjusted to produce a quieter queue rather than to satisfy an integrity-error budget.","Employees could game recorded transitions to remain within the predicted workflow.","Scanner, timestamp, identity, or authorization errors could create misleading residuals.","Risk-stratified audits could concentrate on known concerns and miss novel discrepancy classes.","Residual records could expose sensitive case movements or personnel behavior.","Case-specific model updates could contaminate later custody assessments.","Frequent fallbacks or resynchronizations could eliminate any attention advantage while increasing complexity.","Automation could diffuse responsibility if alerts lack a named owner and acknowledgement requirement.","A model scoped to formal procedure could treat legitimate emergency exceptions as errors or absorb informal deviations that should instead trigger policy review."]},"next_evidence_step":"Pre-register one bounded, offline crossover study using synthetic ledgers or approved records from closed matters. Freeze the predictor, checksum, validity windows, consequence weights, and fallback table. Construct matched sequences containing routine transitions and independently seeded missing scans, duplicated records, delayed returns, wrong locations, unauthorized actors, out-of-order events, seal-state changes, recorder outages, workflow-version changes, and legitimate emergency exceptions. Have blinded reviewers assess randomized sequences through complete-ledger review, the strongest deterministic-rule configuration, and the residual interface. Record discrepancy detection and acknowledgement, protected-class misses, false escalations, unsupported tampering interpretations, full-state reconstruction errors, heartbeat discrimination, audit disagreements, fallback behavior, reviewer time, and model-maintenance effort. Stop on any unexplained protected-class miss or failure to recover the complete ledger. The result may support only another retrospective iteration, not live deployment.","prior_art_status":"UNSEARCHED","diversity_from_prior_proposals":"Earlier proposal 1 addresses analyst attention during examination of digital-device content: it predicts operating-system and application background events so material content-level timeline deviations can be reviewed in context. This proposal addresses physical-evidence custody assurance: it predicts authorized organizational state transitions and routes custody innovations to reconciliation. Its signal is the custody ledger rather than extracted device activity; its scarce channel is quality-assurance attention rather than content examination attention; its downstream action is ledger and physical-state reconciliation rather than forensic interpretation; and its principal failure condition is loss of custody-state integrity rather than omission or overinterpretation of case-content events. It can be adopted independently by a laboratory that does not use residual-first digital timeline review.","revision_record":{"parent_version":null,"progress_targets_addressed":["Created an independently adoptable second candidate for the same archetype-domain cell.","Selected a materially different custody-integrity problem and organizational state-transition causal path.","Specified actors, observables, intervention, archetype structure, authored mechanisms, rivals, authority, safeguards, falsifiers, and bounded evidence.","Explicitly contrasted the candidate with sealed proposal 1."],"conceptual_changes":["Initial version; no parent proposal.","Applied predictive residual processing to physical-evidence custody-state assurance rather than digital-content examination.","Separated residual detection from conclusions about tampering, admissibility, misconduct, or guilt."],"operational_changes":["Specified an append-only full ledger, workflow checksums, verified silence, structured transition residuals, independent quiet-ledger audits, physical-state resynchronization, protected bypasses, and manual fallback.","Bound the first step to frozen-model, offline comparison using synthetic or legally reusable closed records."],"evidence_changes":["Prior art remains unsearched.","Defined a pre-registered comparison against deterministic rules and complete manual reconciliation without asserting an effect size."],"claim_changes":["No novelty, prevalence, demand, or effectiveness claim is made.","The claim is limited to a falsifiable residual-state architecture and conditional value under defined custody, reconstruction, workload, and safety criteria."]}}