{"schema_version":1,"experiment_id":"eoa_inverse_innovation_exp05_complete_proposal_portfolio20_20260803","cell_id":"layer_decay_and_expiration_management__cognitive_science","arm":"COMPLETE_PROPOSAL_PORTFOLIO","candidate_id":"cand_protocol_memory_lifecycle_registry_01","proposal_index":1,"version":0,"title":"Protocol Memory Lifecycle Registry for Evolving Cognitive Tasks","problem":"A cognitive-science laboratory sequentially deposits versions of task instructions, stimulus sets, scoring rules, exclusion criteria, norm tables, and analysis scripts. After a study changes, superseded artifacts can remain searchable and usable without an explicit lifecycle state. A researcher may therefore assemble a task or analysis from mutually incompatible layers, while indiscriminate cleanup could erase the materials needed to reconstruct a published result.","actors":["Principal investigator responsible for the research program","Researchers configuring and administering cognitive tasks","Analysts selecting scoring and exclusion artifacts","Laboratory research-data steward","Internal or external researchers reconstructing a study","Ethics, records, or institutional officials responsible for mandatory preservation"],"observable_state":"For a bounded task family, an artifact inventory shows multiple sequential versions; some lack active, superseded, review-due, archived, quarantined, or preserved-exception labels; some are returned by ordinary search beside the current package; and some have unresolved inbound references from preregistrations, datasets, manuscripts, analysis pipelines, or replication packages. A controlled selection exercise can record whether a researcher chooses the artifact set bound to a specified study version and whether an archived version can be reconstructed.","consequence":"An incompatible combination of instructions, stimuli, scoring, norms, or exclusions can change what construct a task operationalizes or how its observations are interpreted. Conversely, premature destruction can prevent audit, replication, or historical reconstruction of an inference.","affected_objective":"Maintain interpretable and reproducible task-based inference while bounding the active protocol stack and preserving materials required for reconstruction, accountability, ethics, or records obligations.","intervention":"Create a lifecycle registry for each protocol artifact and assembled protocol package. Record identity, creation and supersession dates, study bindings, validation status, access history, inbound dependencies, lifecycle state, and preservation authority. Staleness signals and an age-weighted value score rank items for review but never authorize disposition. A human steward then chooses refresh, retain-active, archive, quarantine, or preserve-by-exception. Archival must retain the package manifest and readers needed for reconstruction; sampled archive-restore drills test that promise. Removal requires an inbound-dependency check, a reversible quarantine window, and a tombstone pointing to any successor. The initial deployment operates in shadow mode and performs no automatic deletion.","structural_mapping":[{"archetype_element":"Accumulated sequential layers","domain_realization":"Successive task instructions, stimuli, scoring rules, exclusion criteria, norm files, software environments, and analysis scripts deposited across studies."},{"archetype_element":"Age and deposition-order index","domain_realization":"Creation, validation, use, and supersession dates recorded for every artifact and assembled protocol package."},{"archetype_element":"Stale layers masquerading as current","domain_realization":"Superseded artifacts remain searchable without a visible status and can be selected for a new task run or analysis."},{"archetype_element":"Differentiated lifecycle states","domain_realization":"Draft, active, review-due, superseded, archived, quarantined, preserved-exception, and destroyed-with-tombstone states."},{"archetype_element":"Dependency and reconstruction check","domain_realization":"Inbound links from preregistrations, study manifests, datasets, manuscripts, pipelines, teaching materials, and replication packages are traced before disposition."},{"archetype_element":"Preservation exceptions","domain_realization":"Named holds protect artifacts required by ethics, institutional records rules, evidentiary needs, published-result reconstruction, or historically significant research."},{"archetype_element":"Reversible cleanup and deletion evidence","domain_realization":"Candidates leave active search through quarantine first; disposition decisions and successor pointers survive separately from removed content."},{"archetype_element":"Revalidation loop","domain_realization":"Protocol owners periodically reassess review dates, dependencies, exceptions, archive readability, and whether an active package still matches its stated construct and study context."}],"mechanism_mapping":[{"mechanism_slug":"stale_layer_detection_dashboard","role":"Provides the identity-resolved inventory and flags artifacts whose content conflicts with the current study manifest, references retired components, lacks an owner, or has been superseded.","counterfactual_removal":"Without it, lifecycle review begins from an incomplete pile and semantically stale artifacts can remain indistinguishable from current ones."},{"mechanism_slug":"age_weighted_value_score","role":"Ranks review candidates using age, recent access, provenance, reconstruction cost, validation status, and misapplication risk; its bands are triage signals only.","counterfactual_removal":"Without it, stewards must inspect an unranked inventory, making bounded review difficult; no safety authority is lost because the score never disposes artifacts."},{"mechanism_slug":"retention_schedule","role":"Maps artifact classes to review periods, minimum preservation needs, disposition options, and named ethics or institutional overrides.","counterfactual_removal":"Without it, retention decisions become case-by-case judgments and preservation exceptions can become implicit or inconsistently applied."},{"mechanism_slug":"dependency_safe_delete_check","role":"Blocks removal when a live study, dataset, manuscript, pipeline, replication package, or mandatory reconstruction path still references the artifact.","counterfactual_removal":"Without it, a low-value or old artifact could be removed even though it remains load-bearing for scientific reconstruction."},{"mechanism_slug":"lifecycle_storage_tiering_policy","role":"Moves superseded but reconstructively useful packages out of active search into slower archival tiers while retaining their manifests and identities.","counterfactual_removal":"Without it, the design collapses toward either keeping every artifact active or destroying material that is rarely accessed but still required."},{"mechanism_slug":"archive_restore_test","role":"Periodically retrieves sampled archived packages across age and format bands, opens their contents, and reconnects them to a sandbox reconstruction workflow.","counterfactual_removal":"Without it, archived status supplies no operational evidence that the protocol can still be read and reconstructed."},{"mechanism_slug":"soft_delete_quarantine_window","role":"Hides an approved retirement candidate from ordinary use while keeping it recoverable for a time-boxed period determined by reconstruction impact.","counterfactual_removal":"Without it, an erroneous disposition becomes immediately irreversible."},{"mechanism_slug":"tombstone_or_deletion_marker","role":"Preserves the retired artifact's identity, disposition reason, date, and successor pointer after content leaves active use.","counterfactual_removal":"Without it, later searches cannot distinguish deletion from nonexistence, and surviving references may resolve ambiguously."}],"causal_chain":["Sequential studies deposit new protocol artifacts faster than superseded artifacts receive explicit retirement decisions.","Unlabeled older artifacts remain visible beside current artifacts, while their reconstruction dependencies are not readily inspectable.","The registry binds every artifact to an identity, temporal position, study package, lifecycle state, and owner.","Semantic staleness signals and age-weighted value rank review candidates without changing or deleting them.","The retention schedule and exception register determine which candidates require preservation or human escalation.","The inbound-dependency gate prevents disposition of artifacts still needed by live work or reconstruction paths.","Approved candidates leave active search through archival or reversible quarantine, with tombstones preserving identity and successor resolution.","Archive-restore drills and periodic exception reviews detect failures at the far end of the lifecycle and return affected packages for repair.","The active stack becomes bounded and status-explicit while reconstructively necessary protocol history remains recoverable."],"baseline":"A shared drive or repository preserves folders and commit history, researchers use filenames such as final or latest plus informal README guidance, and cleanup occurs when a person notices clutter. Version history records change but does not consistently assign lifecycle state, test semantic staleness, expose inbound scientific dependencies, govern exceptions, or prove archived packages reconstructable.","nearest_rivals":["An immutable version-control repository with a single current-version pointer","A study manifest or container that freezes the exact artifacts used for each experiment","Periodic manual repository cleanup guided by owners and README files","A conventional research-data retention policy applied uniformly to whole projects"],"remaining_contrastive_claim":"The proposal's distinct testable contribution is lifecycle governance across individual protocol layers: it separates active authority from historical recoverability, requires dependency-gated and reversible disposition, and tests reconstruction after archival. A current-version pointer or frozen study package can identify a version, but does not by itself retire misleading layers, manage exceptions, or validate restoration.","authority_safety":{"decision_authority":"The principal investigator and designated research-data steward jointly approve lifecycle rules; artifact owners review scientific dependencies; institutional ethics or records authorities control applicable preservation overrides. No ranking model has disposition authority.","authorized_first_step":"Build a read-only shadow registry for one completed task family, label provisional lifecycle states, trace visible dependencies, and simulate archive restoration without moving or deleting source artifacts.","excluded_actions":["Automatic hard deletion","Deletion or relocation of participant data","Removal of materials under ethics, records, legal, or evidentiary obligations","Changing published analyses, results, or task definitions","Treating an age-weighted score as a scientific-validity judgment","Breaking a live study binding or reconstruction path","Declaring an artifact obsolete solely because it is old"],"halt_rollback":"Halt the pilot if the registry misidentifies artifact identity, omits a known dependency, exposes restricted material, or cannot restore a sampled package. Because the first step is read-only, rollback consists of disabling the shadow index and deleting only its derived labels after retaining a record of the pilot error; source artifacts remain unchanged."},"negative_tests":{"strongest_counterevidence":"If study manifests or immutable package identifiers already determine artifact selection, reconstruction succeeds from existing repositories, and observed mismatches arise from task-design ambiguity rather than stale artifact discovery, the lifecycle intervention addresses little of the consequential problem.","problem_falsifier":"For the selected task family, every artifact is uniquely bound to a study and visibly labeled, no superseded artifact appears as current in ordinary retrieval, all inbound reconstruction dependencies are known, and blinded users consistently select the specified package without additional lifecycle information.","intervention_falsifier":"Compared with a lightweight current-version manifest, the shadow lifecycle registry does not reduce wrong-package selections or improve successful reconstruction in the bounded exercise, or it introduces unresolved dependencies, misleading status labels, or failed restores that its gates do not catch.","risks":["Composite scores can create false precision and embed disputed judgments about scientific value.","Incomplete dependency tracing can falsely classify a load-bearing artifact as an orphan.","Staleness flags can be mistaken for evidence that a construct or result is invalid.","Lifecycle metadata can reveal restricted study or participant information if scope controls fail.","Archival can impose retrieval delays during replication or investigation.","Preservation exceptions can become permanent by inertia and recreate the original accumulation problem.","Quarantined content still exists and may conflict with a binding destruction obligation.","Stewardship effort can exceed the value of managing very small or already well-bound artifact collections."]},"next_evidence_step":"In one consenting laboratory, select one completed cognitive-task family containing three to five documented revisions and exclude participant data. Construct the read-only registry, have the artifact owner verify identities and inbound links, and sample one package from each available age or format band for sandbox restoration. Then give blinded researchers a fixed set of study-reconstruction prompts and compare package selection using the existing repository interface versus the shadow lifecycle view. Record selection correctness, unresolved dependencies, restoration success, elapsed review effort, false staleness flags, and any safeguard-triggered halt. Do not move, quarantine, or delete source artifacts during this evidence step.","prior_art_status":"UNSEARCHED","diversity_from_prior_proposals":"No prior experiment candidates were consulted; this proposal is specified solely as lifecycle control of sequential cognitive-task protocol artifacts under the sealed input.","revision_record":{"parent_version":null,"progress_targets_addressed":["Initial complete specification of the inferred domain problem","Explicit preservation of the archetype's accumulation, expiry, exception, dependency, reversibility, and restoration structure","Operational authority boundaries and falsifiable first evidence"],"conceptual_changes":["Initial version; no parent proposal"],"operational_changes":["Initial version; defines a read-only one-task-family shadow pilot with no automatic disposition"],"evidence_changes":["Initial version; specifies observable selection, dependency, and restoration checks without effect-size claims"],"claim_changes":["Initial version; limits the contrastive claim to lifecycle governance beyond version identification and frozen packaging","Makes no claim of novelty, prevalence, demand, or effect size"]}}