{"schema_version":1,"experiment_id":"eoa_inverse_innovation_exp05_complete_proposal_portfolio20_20260803","cell_id":"layer_decay_and_expiration_management__psychology","arm":"COMPLETE_PROPOSAL_PORTFOLIO","candidate_id":"psychology_protocol_exception_sunset_board_v0","proposal_index":5,"version":0,"title":"Protocol-Exception Sunset Board for Behavioral Experiments","problem":"Behavioral experiments can acquire successive procedural exceptions after the approved protocol is written: temporary equipment substitutions, remote-administration adaptations, revised prompts, scheduling workarounds, participant accommodations, revised exclusion rules, and clarifications issued after unexpected cases. These layers may be scattered across messages, meeting notes, scripts, and staff memory. An exception created for one period or condition can continue shaping later sessions after its trigger has disappeared or a newer rule has superseded it. Deleting old exceptions is also unsafe because completed data batches, ethics decisions, deviations, and replication attempts may depend on reconstructing the exact procedure in force at the time.","actors":["Principal investigator","Behavioral-laboratory manager","Experimenter administering study sessions","Research ethics or human-subjects reviewer","Research data manager or reproducibility reviewer"],"observable_state":"The operative procedure is an unresolved stack of a base protocol plus later exceptions whose scope, approval, effective dates, owners, supersession relationships, and affected data batches are not consistently recorded. Observable indicators include different staff applying different instructions to equivalent sessions, temporary workarounds appearing in current run sheets after their conditions end, contradictory exclusion or prompting rules, amendments with departed owners, scripts that cannot be linked to an approved procedural state, and inability to reconstruct which exception set governed a completed batch.","consequence":"Stale or conflicting procedural layers can produce unintended variation in administration, inconsistent participant treatment, ambiguous exclusions, and uncertainty about which procedure generated a dataset. Aggressive cleanup can instead erase the lineage needed for ethics review, deviation investigation, replication, or interpretation of earlier batches.","affected_objective":"Present experimenters with one explicit, currently authorized procedure for each session context while preserving the exact exception stack needed to reconstruct prior administrations and data batches.","intervention":"Install a Protocol-Exception Sunset Board that treats each deviation, waiver, clarification, accommodation rule, and temporary workaround as an individually governed layer rather than editing it invisibly into a monolithic procedure. Every exception records its trigger context, scope, owner, approval basis, effective date, review lease, affected protocol components, incompatible rules, dependent sessions or data batches, successor, and lifecycle state: active, review-due, superseded, quarantined, or archived. Before a session, the board compiles a read-only run sheet from the approved base protocol plus only the active exceptions whose trigger conditions match that session. Temporary exceptions receive review leases at authorization; expiry removes them from automatic compilation only after the responsible authority confirms renewal, supersession, or retirement. Ethics- and safety-linked exceptions cannot expire automatically. Superseded layers leave markers, proposed removals enter reversible quarantine, and archived exception sets remain restorable by session or batch identifier.","structural_mapping":[{"archetype_element":"Sequential deposits","domain_realization":"Clarifications, deviations, accommodations, equipment substitutions, and temporary workarounds accumulate after the base experimental protocol is established."},{"archetype_element":"Changed active usefulness","domain_realization":"An exception can cease to apply when equipment is restored, administration mode changes, recruitment ends, a successor rule is approved, or its triggering condition disappears."},{"archetype_element":"Stale layers remaining active","domain_realization":"Old exceptions continue appearing in scripts, instructions, or staff practice because they lack explicit expiry and supersession states."},{"archetype_element":"Inventory and identity map","domain_realization":"Each exception receives an identity linked to its source approval, affected protocol components, trigger conditions, and dependent sessions or batches."},{"archetype_element":"Expiration trigger","domain_realization":"A review lease makes a temporary exception review-due at a predefined point without automatically reversing an authorized safety or ethics control."},{"archetype_element":"Dependency and reconstruction check","domain_realization":"Retirement checks whether completed sessions, data batches, analyses, deviation reviews, or replication packages still reference the exception."},{"archetype_element":"Differentiated disposition","domain_realization":"Exceptions can be renewed, narrowed, superseded, quarantined, archived, or preserved by exception rather than merely retained or deleted."},{"archetype_element":"Current-state compilation","domain_realization":"A session run sheet includes the base protocol and only the currently active, context-matching exception layers."},{"archetype_element":"Audit-preserving retirement","domain_realization":"Disposition markers and restorable archives preserve what governed earlier sessions after an exception leaves current use."}],"mechanism_mapping":[{"mechanism_slug":"stale_layer_detection_dashboard","role":"Inventories exception layers and flags expired leases, absent trigger contexts, departed owners, contradictory rules, missing approvals, broken component links, and unrecorded successors for review.","counterfactual_removal":"Without a resolved inventory and stale-context detector, obsolete exceptions remain dispersed across scripts and communications and cannot be reliably excluded from current procedure compilation."},{"mechanism_slug":"time_to_live_ttl_policy","role":"Assigns temporary exceptions an authorization-specific review lease; expiry changes the layer to review-due and prevents silent indefinite continuation while requiring human disposition.","counterfactual_removal":"Without the precommitted review trigger, workarounds created for transient conditions can become permanent through inertia."},{"mechanism_slug":"age_weighted_value_score","role":"Ranks review-due exceptions using age, recent invocation, current trigger prevalence, approval significance, reconstruction value, and conflict risk; it cannot authorize retirement or procedural change.","counterfactual_removal":"Without a comparable review ranking, large exception stacks must be examined indiscriminately or reduced using age alone, obscuring old but consequential layers."},{"mechanism_slug":"retention_schedule","role":"Maps exception classes to review periods, required retention, disposition pathways, and ethics, safety, complaint, or reproducibility preservation overrides.","counterfactual_removal":"Without the policy matrix, equivalent exceptions could receive inconsistent lifetimes and protected procedural evidence could be removed ad hoc."},{"mechanism_slug":"dependency_safe_delete_check","role":"Traces inbound references from session records, batch manifests, analysis decisions, ethics reviews, deviation reports, and replication packages before an exception can leave recoverable storage.","counterfactual_removal":"Without the inbound-reference gate, cleanup could make a completed dataset's generating procedure impossible to reconstruct."},{"mechanism_slug":"soft_delete_quarantine_window","role":"Removes an approved retired exception from current compilation while retaining it in recoverable quarantine for a period based on affected sessions, unresolved references, and reversal impact.","counterfactual_removal":"Without quarantine, a mistaken retirement could alter upcoming administration before a missed dependency or trigger condition is discovered."},{"mechanism_slug":"tombstone_or_deletion_marker","role":"Leaves a marker at the exception identifier recording its disposition, date, authority, reason, successor, and archive location so scripts and batch references resolve unambiguously.","counterfactual_removal":"Without a marker, a missing exception could be mistaken for a never-approved rule, and stale copies could silently reintroduce a superseded procedure."},{"mechanism_slug":"archive_restore_test","role":"Periodically restores sampled historical exception sets and recompiles the run sheet that governed a selected session or batch, checking fidelity and dependency reconnection.","counterfactual_removal":"Without an end-to-end reconstruction test, archived layers could become unreadable or detached from the sessions they are supposed to explain."}],"causal_chain":["Operational contingencies deposit successive exception layers over the approved base experimental protocol.","The board identifies each layer and records its scope, trigger, authority, age, dependencies, conflicts, and lifecycle state.","Review leases and stale-context signals surface temporary exceptions whose continuing authority requires reassessment.","Before each session, context matching selects the approved base protocol and applicable active exceptions to compile one run sheet.","Authorized reviewers renew, narrow, supersede, quarantine, or archive review-due layers after checking ethics, safety, and data dependencies.","Retired or superseded exceptions stop shaping new sessions, reducing conflicts among accumulated procedural rules.","Session and batch manifests record the exact base and exception identifiers used.","Markers, quarantine, archives, and reconstruction drills preserve the ability to explain completed administrations and data even after their exceptions expire."],"baseline":"Laboratories commonly maintain an approved protocol, procedural manual, scripts, amendment files, and informal clarifications. Staff determine the operative combination through training, local copies, messages, or memory. Version control may preserve whole files, but temporary exceptions are not necessarily represented as separately expiring objects, and the exact rule stack applied to each session or batch may remain implicit.","nearest_rivals":["Whole-document version control that preserves successive protocols but does not independently expire conditional exceptions or select them by session context","Standard operating procedure revision that periodically publishes a new consolidated document while leaving the status and dependencies of temporary deviations implicit","Protocol preregistration that establishes an initial plan but does not manage the lifecycle of operational exceptions arising during execution","Experimenter checklists that improve adherence to a chosen procedure but cannot determine which accumulated exception layers currently belong on the checklist","Deviation logs that record departures after they occur without compiling the authorized rule set for future sessions or retiring obsolete deviations"],"remaining_contrastive_claim":"The proposal's testable contrast is lifecycle governance for conditional protocol exceptions: individually identified layers receive review leases, are selected by runtime context, cannot retire while referenced, and remain reconstructable after leaving current administration. Whole-file versioning, SOP revision, preregistration, checklists, and deviation logging do not jointly produce that causal path.","authority_safety":{"decision_authority":"The principal investigator may renew or retire ordinary operational exceptions within the approved protocol. Any exception affecting consent, participant risk, eligibility protections, deception, compensation, or other ethics-controlled terms requires the existing human-subjects authority. Laboratory staff may report conflicts and use compiled run sheets but cannot approve procedural changes. The board has no authority to determine scientific or ethical acceptability.","authorized_first_step":"Model one completed synthetic experiment with a base protocol and seeded exception history, then test read-only compilation and historical reconstruction. No live protocol, participant session, ethics record, script, or dataset is changed.","excluded_actions":["Automatically approving, renewing, or retiring a protocol exception","Allowing lease expiry to disable a safety, consent, accommodation, or ethics requirement","Rewriting signed protocols, consent materials, deviation reports, or historical session records","Using an age-weighted score as evidence that a procedural rule is scientifically invalid","Compiling an unapproved exception into a participant-facing procedure","Deleting a layer referenced by a completed session, data batch, complaint, audit, or replication package","Replacing investigator or ethics review with automated context matching","Applying a changed run sheet to an ongoing study without required authorization"],"halt_rollback":"Halt if the compiled run sheet omits an applicable protected exception, includes a superseded or unapproved rule, cannot trace every included layer to authority, or fails to reconstruct a seeded historical batch. Disable compilation, restore all quarantined test layers to their prior states, preserve the audit output, and return to the authorized baseline procedure."},"negative_tests":{"strongest_counterevidence":"Procedural exceptions may encode tacit expertise or rare participant protections that cannot be captured by explicit trigger metadata. Administration inconsistencies may arise primarily from training or adherence failures, so lifecycle machinery could add bureaucracy without addressing the causal problem.","problem_falsifier":"A bounded audit finds no unresolved exception stack: staff consistently identify the same authorized procedure for each context, temporary rules have explicit end states, every session records its exact governing version, and all sampled batches can be reconstructed without missing or contradictory layers.","intervention_falsifier":"In simulated administration and reconstruction tasks, the board does not reduce inclusion of obsolete rules or disagreement about the current procedure, or it increases omission of valid accommodations, unapproved compilations, review burden, execution time, or reconstruction failures beyond predeclared limits.","risks":["Explicit trigger metadata may fail to represent ambiguous real-world situations.","Experimenters may overtrust a compiled run sheet and stop applying necessary judgment.","A lease could be mistaken for permission to disable a protected requirement.","Reviewers may renew every exception mechanically, preserving the stack indefinitely.","Frequent compilation changes could introduce new inconsistency across sessions.","Dependency mapping may miss informal analytical or interpretive reliance on an exception.","Detailed exception histories could expose sensitive deviations or participant circumstances.","Archiving executable scripts without their software environment may not preserve reconstruction.","The board could normalize excessive procedural deviation by making exceptions easier to administer."]},"next_evidence_step":"Construct a preregistered synthetic protocol containing a base procedure and seeded layers for equipment failure, remote administration, wording clarification, accommodation, revised exclusion handling, ethics-controlled consent, supersession, and batch-specific deviation. Randomly assign trained reviewers to the representative document-and-message baseline or a read-only Sunset Board. Measure correct current run-sheet composition across session contexts, inclusion of obsolete or unapproved rules, omission of protected exceptions, reviewer agreement, time, dependency-detection accuracy, and reconstruction of the exact procedure for historical batches. Perform one quarantine rollback and one archived run-sheet restoration. Do not involve participants or alter any live study.","prior_art_status":"UNSEARCHED","diversity_from_prior_proposals":"Proposal 1 governs clinician-authored hypotheses about an individual patient; proposal 5 instead governs conditional rules that alter experimental administration and never evaluates patient evidence. Proposal 2 manages participant-owned behavior-change commitments under an attention budget; proposal 5 has no personal goal portfolio, reminder capacity, or participant-controlled sunset decision. Proposal 3 manages norm and scoring packages through a context-sensitive scoring gate; proposal 5 operates upstream of scoring by compiling the procedure used to generate behavioral observations, and its layers are exceptions rather than calibration assets. Proposal 4 manages person-linked histories of prior paradigm exposure for enrollment eligibility; proposal 5 manages protocol-linked operational exceptions applied during study execution and does not retain cross-study participant histories. Its causal path—from accumulated procedural exceptions through context compilation and authorized sunset to uniform current administration with batch reconstruction—is independently adoptable as laboratory protocol governance.","revision_record":{"parent_version":null,"progress_targets_addressed":["Generate proposal index 5 as one complete candidate","Address a problem materially different from proposals 1 through 4","Use an independently adoptable intervention and distinct causal path","Preserve the layer-decay and expiration structure and authored mechanisms","Explain diversity from every earlier sealed proposal","Specify operational authority, safeguards, rivals, falsifiers, and bounded evidence"],"conceptual_changes":[],"operational_changes":[],"evidence_changes":[],"claim_changes":[]}}