{"schema_version":1,"experiment_id":"eoa_inverse_innovation_exp03_full320_20260801","cell_id":"layer_decay_and_expiration_management__neuroscience","trajectory_id":"R","attempt_index":0,"archetype_slug":"layer_decay_and_expiration_management","domain_slug":"neuroscience","decision":"CANDIDATE","problem_id":"longitudinal_neural_recording_version_accumulation","causal_lever_id":"state_gated_reversible_neural_data_retirement","proposal":{"problem":"Longitudinal electrophysiology programs accumulate raw recordings, spike-sorting outputs, annotations, and reprocessed versions faster than obsolete or low-access layers are retired. Old derivatives may remain discoverable as if authoritative, while indiscriminate deletion can destroy reproducibility, participant-governance evidence, or material newly useful to future analyses.","actors_substrate":["human or animal research participants","principal investigators and neuroscience data stewards","analysts and downstream reusers","raw neural recordings, event metadata, spike sorts, annotations, and derived datasets","institutional storage, consent, ethics, and records-governance systems"],"observable_state":"Multiple age-ordered versions and session deposits coexist without consistent lifecycle states; superseded derivatives appear in normal search, hot storage grows, ownership is unclear, and deletion is deferred because dependencies and restoreability are unknown.","consequence":"Researchers can analyze stale versions, spend attention and fast storage on inactive material, or lose lineage and reconstruction capacity through improvised cleanup.","affected_objective":"HYPOTHESIS: improve authoritative-dataset identification, reproducibility, retrieval effort, and bounded storage cost without weakening participant protections or recovery.","structural_mapping":[{"archetype_element":"sequential deposits","domain_realization":"Recording sessions and successive preprocessing, spike-sorting, annotation, and correction outputs accumulate over experiment time.","claim_kind":"INFERENCE"},{"archetype_element":"old layers remain active by default","domain_realization":"Superseded neural-data derivatives remain searchable or referenced without an explicit non-authoritative state.","claim_kind":"HYPOTHESIS"},{"archetype_element":"retention value differs by layer","domain_realization":"Unique raw signals, regenerable intermediates, validated releases, and temporary scratch outputs have different reconstruction and governance value.","claim_kind":"INFERENCE"},{"archetype_element":"deletion can break reconstruction","domain_realization":"Removing source signals, parameters, event metadata, or lineage links can prevent reproduction of published neural analyses.","claim_kind":"INFERENCE"},{"archetype_element":"managed lifecycle","domain_realization":"Classify, label, tier, archive, quarantine, restore-test, and only then destroy eligible deposits.","claim_kind":"HYPOTHESIS"}],"component_map":[{"component":"layer_inventory_and_identity_map","status":"direct","domain_realization":"Registry of sessions and raw, processed, annotated, and released versions."},{"component":"deposition_order_and_age_index","status":"direct","domain_realization":"Acquisition and derivation timestamps plus version order."},{"component":"retention_policy_matrix","status":"adapted","domain_realization":"Rules by raw data, validated release, regenerable intermediate, and scratch class."},{"component":"decay_function_or_aging_rule","status":"adapted","domain_realization":"Age reduces presumed operational value but never overrides unique-source or governance value."},{"component":"expiration_trigger_definition","status":"adapted","domain_realization":"Inactivity, supersession, project closure, or scratch TTL triggers review or demotion."},{"component":"layer_value_and_risk_score","status":"adapted","domain_realization":"Score access, uniqueness, reproducibility, sensitivity, regeneration cost, and misleading-version risk."},{"component":"dependency_and_reconstruction_check","status":"direct","domain_realization":"Trace publications, pipelines, releases, and lineage before retirement."},{"component":"pruning_or_archival_pathway","status":"direct","domain_realization":"Refresh, archive, compact, quarantine, or approved destruction."},{"component":"preservation_exception_register","status":"direct","domain_realization":"Named exceptions for unique cohorts, published evidence, disputes, or historical value."},{"component":"deletion_audit_and_rollback_window","status":"direct","domain_realization":"Recorded soft deletion with a time-bounded restore period."},{"component":"resource_budget_and_storage_tier","status":"direct","domain_realization":"Fast, nearline, cold, and archival capacity budgets."},{"component":"review_cadence_and_revalidation_loop","status":"direct","domain_realization":"Periodic owner, exception, and restore revalidation."},{"component":"legal_hold_or_compliance_override","status":"adapted","domain_realization":"Consent, ethics, investigation, funder, or institutional-record override."},{"component":"hot_warm_cold_access_state","status":"direct","domain_realization":"Access tier recorded for every deposit."},{"component":"orphan_layer_detection","status":"direct","domain_realization":"Identify derivatives with no live lineage, release, publication, or pipeline reference."},{"component":"supersession_marker","status":"direct","domain_realization":"Non-authoritative version points to its validated successor."},{"component":"layer_compaction_candidate_flag","status":"adapted","domain_realization":"Flag small files or regenerable intermediates for lossless packaging or deduplication."},{"component":"rehydration_or_restore_procedure","status":"direct","domain_realization":"Documented retrieval, integrity validation, parsing, and lineage reconnection."},{"component":"deletion_impact_estimate","status":"direct","domain_realization":"Estimate affected analyses, users, reconstruction cost, and participant-governance impact."},{"component":"stale_context_detector","status":"adapted","domain_realization":"Detect outputs tied to obsolete pipelines, annotations, references, or unit identities."}],"mechanism_dispositions":[{"slug":"age_weighted_value_score","disposition":"selected_supporting","contribution_type":"OPERATIONAL","adaptation_or_rejection":"Use only to rank review; uniqueness, consent, and dependencies are vetoes because scientific value need not decay with age.","counterfactual_removal":"Review becomes less scalable, but class rules and human review can still govern disposition."},{"slug":"archive_restore_test","disposition":"selected_load_bearing","contribution_type":"SAFETY_GUARDRAIL","adaptation_or_rejection":"Sample archived neural datasets across formats and ages; verify integrity, parsing, metadata, and lineage reconnection.","counterfactual_removal":"Archival could silently become irreversible loss, hard-gating safe demotion."},{"slug":"cache_eviction_rule","disposition":"considered_rejected","contribution_type":"NONE","adaptation_or_rejection":"Access recency alone is unsafe for scientifically unique data; ordinary scheduled tiering is sufficient.","counterfactual_removal":"No material causal loss because capacity pressure must not choose destruction or authoritative status."},{"slug":"dependency_safe_delete_check","disposition":"selected_load_bearing","contribution_type":"SAFETY_GUARDRAIL","adaptation_or_rejection":"Gate deletion on live lineage, publication, workflow, and external-use checks; incomplete visibility blocks deletion.","counterfactual_removal":"Cleanup could break reproducibility or active analyses, making destructive disposition unacceptable."},{"slug":"lifecycle_storage_tiering_policy","disposition":"selected_load_bearing","contribution_type":"CORE_CAUSAL","adaptation_or_rejection":"Demote low-access deposits while preserving identity and retrieval; compact only losslessly or when regeneration is proven.","counterfactual_removal":"The intervention cannot bound expensive active storage while retaining scientifically valuable history."},{"slug":"log_rotation_and_cleanup_job","disposition":"considered_rejected","contribution_type":"NONE","adaptation_or_rejection":"Session datasets are semantically heterogeneous; generational position cannot safely determine disposal.","counterfactual_removal":"Lifecycle classification and tiering still operate without blind rotation."},{"slug":"retention_schedule","disposition":"selected_load_bearing","contribution_type":"CORE_CAUSAL","adaptation_or_rejection":"Map neuroscience data classes to minimums, maximums, disposition, consent constraints, and named exceptions.","counterfactual_removal":"Disposition lacks legitimate, repeatable class-level authority."},{"slug":"soft_delete_quarantine_window","disposition":"selected_load_bearing","contribution_type":"SAFETY_GUARDRAIL","adaptation_or_rejection":"Quarantine eligible layers before destruction; shorten or omit only where a binding destruction requirement forbids continued retention.","counterfactual_removal":"Mistaken cleanup becomes immediately irreversible."},{"slug":"stale_layer_detection_dashboard","disposition":"selected_load_bearing","contribution_type":"CORE_CAUSAL","adaptation_or_rejection":"Expose inventory, ownership, supersession, pipeline validity, and lifecycle state without acting automatically.","counterfactual_removal":"Stale versions remain indistinguishable from current authoritative outputs."},{"slug":"time_to_live_ttl_policy","disposition":"selected_supporting","contribution_type":"OPERATIONAL","adaptation_or_rejection":"Restrict TTL to scratch outputs and explicitly temporary derivatives; expiry means review or quarantine, not raw-data destruction.","counterfactual_removal":"Temporary accumulation requires scheduled sweeps but the main lifecycle remains viable."},{"slug":"tombstone_or_deletion_marker","disposition":"selected_supporting","contribution_type":"SAFETY_GUARDRAIL","adaptation_or_rejection":"Leave a minimal provenance-safe marker identifying deletion or successor without retaining prohibited participant data.","counterfactual_removal":"References can become ambiguous and stale replicas or indexes may revive retired versions."}],"causal_chain":["Inventory resolves deposits, identities, owners, age, lineage, and authority state.","Class policy and staleness signals separate active, superseded, archival, and review candidates.","Dependency and governance gates prevent retirement of load-bearing or protected material.","Tiering and lossless compaction reduce active resource burden while retaining recoverability.","Quarantine, tombstones, audit, and tested restoration make approved retirement accountable and reversible.","Periodic revalidation prevents hidden permanent exceptions and archive decay."],"baseline":"Projects retain most raw and derived files in their existing locations; analysts rely on folder names, local conventions, and ad hoc cleanup, while uncertain dependencies encourage indefinite retention.","nearest_rival":"Improve cataloging and version labels while retaining every deposit on its current tier. This addresses discovery errors but not active-storage growth, archive viability, or accountable retirement.","authority_safety":{"affected_parties":["research participants or animal-study stakeholders","current investigators and data stewards","authors, analysts, and downstream reusers","institutional ethics, privacy, and records officers"],"decision_authority":"The accountable principal investigator and institutional data steward approve policy; ethics, consent, privacy, records, and investigation authorities can veto disposition.","authorized_first_step":"Run a 90-day, read-only pilot on one completed study: inventory deposits, label superseded derivatives, simulate tiers and deletion eligibility, and restore-test a sample; permit reversible demotion of copied non-sensitive derivatives only, with no source-data destruction.","excluded_actions":["automatic deletion of raw or uniquely irreproducible neural signals","destruction under unresolved consent, ethics, investigation, publication, or dependency status","lossy compaction that changes scientific measurements without separate validation","treating age or low access as sufficient evidence of low scientific value"],"halt_rollback":"Halt demotion or cleanup on missing lineage, contested authority, restore failure, integrity mismatch, unexpected downstream breakage, or participant-governance conflict; restore quarantined items and retain the audit record."}},"negative_tests":{"strongest_counterevidence":"Old neural recordings can gain value when new algorithms or questions emerge, and storage may be cheaper than proving safe deletion; if uniqueness dominates cost and search is already governed, age-based retirement could reduce future discovery more than it helps.","analogy_break":"Unlike routine administrative layers, a neural recording's scientific value is not reliably monotonic with age, and the underlying biological event cannot be recreated. Therefore decay scores and TTLs cannot govern unique raw signals.","failure_condition":"The approach fails if identities and dependencies cannot be resolved well enough to distinguish regenerable derivatives from unique evidence, or if binding governance rules require incompatible retention outcomes.","problem_falsifier":"A complete inventory shows bounded storage, negligible stale-version use, reliable authoritative-version labels, known dependencies, and no material retrieval, cost, compliance, or maintenance burden attributable to accumulated layers.","intervention_falsifier":"In the bounded pilot, lifecycle labeling and simulated tiering do not improve authoritative-version selection, hot-tier demand, or retrieval effort, or any demoted sample fails restoration, integrity, lineage reconnection, or governance review.","risks":["age-weighted scores create false precision and bias against historically valuable data","cold-tier latency impedes unexpected reanalysis","invisible external dependencies make orphan findings wrong","quarantine conflicts with promised destruction deadlines","tombstones leak sensitive metadata or themselves accumulate","lossless-looking compaction alters formats or impairs future tooling"]},"null_rationale":null,"classification":{"candidate_kind":"DOMAIN_TRANSFER","prior_art_status":"UNSEARCHED","evidence_maturity":"HYPOTHESIS"},"revision_change_log":{"revision_kind":"ORIGINAL","prior_problem_id":null,"prior_causal_lever_id":null,"problem_changed":false,"causal_lever_changed":false,"conceptual_changes":[],"operational_changes":[],"repairs_addressed":[]},"confidence":0.82,"generator_notes":"Closed-book structural transfer. The candidate concerns lifecycle governance of neuroscience research artifacts, not biological forgetting or activation decay; empirical prevalence and effect sizes remain untested."}