{"actors":["Collection assessment librarian who runs circulation and holdings analyses","Branch librarians who interpret local collection context","Collection-development committee that recommends transfers, storage, or withdrawal","Library director or delegated collection authority who approves consequential action","Catalog and circulation systems analyst who maintains the analysis log"],"affected_objective":"Make evidence-based collection-management decisions without treating the most striking pattern found across many dashboard slices as a confirmed change in local use.","arm":"COMMON_P1","authority_safety":{"authorized_first_step":"The collection-development committee may authorize a read-only shadow review using historical, de-identified aggregate records; outputs remain advisory and cannot change item status.","decision_authority":"Only the library director or formally delegated collection authority may approve transfers, storage, or withdrawals after documented confirmation and branch review.","excluded_actions":["No automatic withdrawal, relocation, acquisition cancellation, or budget reallocation from an exploratory result","No release of patron-level borrowing histories or small-cell reports","No retroactive redefinition of the claim family after results are viewed","No reuse of confirmation-period data to tune the discovery rule","No inference that low recorded circulation means low cultural, educational, preservation, or community value"],"halt_rollback":"Halt the pilot if attempted looks cannot be reconstructed, confirmation data are contaminated, privacy suppression fails, or staff use exploratory labels to justify action. Quarantine affected reports, restore any prematurely changed recommendation to pending, and rerun only after the family and data boundary are fixed."},"baseline":"Staff inspect circulation dashboards across branches, classification bands, metrics, and time windows, then elevate an unusually low-use slice for weeding review without retaining a complete inventory of other slices examined or distinguishing discovery from confirmation.","candidate_id":"multiple_testing_discipline__library_information_science__COMMON_P1","causal_chain":["A collection review exposes staff to many branch-by-classification-by-metric-by-window comparisons.","Ordinary per-comparison thresholds give chance variation repeated opportunities to produce an apparently exceptional decline.","The dashboard preserves the selected slice more visibly than null or contradictory slices.","Reviewers interpret the selected pattern as a stand-alone finding and may recommend consequential collection changes.","A cycle-level claim registry makes every attempted comparison and analytic choice part of the evidence context.","A predeclared error-risk policy applies a familywise rule to withdrawal candidates while retaining nonqualifying patterns only as exploratory leads.","Status labels prevent exploratory and adjusted-discovery results from entering the action queue.","A targeted test on untouched subsequent-period data can promote a lead to confirmed status only under locked criteria.","The authorized decision-maker receives the confirmation record plus branch context before considering any collection change."],"cell_id":"multiple_testing_discipline__library_information_science","consequence":"A chance-selected circulation pattern can trigger avoidable withdrawal, relocation, acquisition cuts, or reputational claims about community interests, while null and conflicting analyses remain invisible.","diversity_from_prior_proposals":"The candidate is grounded specifically in collection-weeding analytics and its reversible-to-irreversible decision boundary; runtime isolation precludes comparison with prior proposals.","experiment_id":"eoa_inverse_innovation_exp13_second_slot_policy60_20260806","intervention":"Create a collection-review discovery register that defines each quarterly family as every branch, classification band, circulation metric, eligibility filter, and time window inspected. Log all attempted looks automatically, lock the family before review, and label outputs exploratory by default. Apply a predeclared Holm familywise rule to any lead proposed for withdrawal review; then require the same directional claim to pass locked criteria on an untouched subsequent circulation period and receive branch-context review before it becomes decision-eligible. Preserve null, rejected, and confirmed claims with their analytic specifications.","mechanism_mapping":[{"counterfactual_removal":"Without the registry, failed and alternative dashboard slices can disappear, making the selected result look like a single planned comparison.","mechanism_slug":"claim_registry","role":"Records the complete family of attempted collection-use claims, specifications, owners, and statuses."},{"counterfactual_removal":"Without adjusted family-level thresholds, the intervention would document search but leave the selected finding subject to ordinary single-comparison interpretation.","mechanism_slug":"bonferroni_like_correction","role":"Implements a Holm familywise adjustment for leads that could advance toward withdrawal review."},{"counterfactual_removal":"Without untouched data, the same circulation variation used to select a lead would also be used to certify it.","mechanism_slug":"holdout_validation","role":"Reserves a subsequent circulation period for a locked, targeted confirmation test."},{"counterfactual_removal":"Without a follow-up gate, an adjusted discovery could be treated as sufficient authorization for a consequential collection change.","mechanism_slug":"confirmatory_follow_up","role":"Requires prospective confirmation and branch interpretation before decision eligibility."},{"counterfactual_removal":"Without ordered labels, attractive exploratory results could be rhetorically promoted despite failing the required evidentiary stage.","mechanism_slug":"claim_registry","role":"Enforces exploratory, adjusted-discovery, confirmed, rejected, and decision-eligible statuses in reports and queues."}],"nearest_rivals":["A single low-circulation threshold, which standardizes item review but does not account for the many slices searched before choosing a target","A reproducibility protocol, which can rerun the dashboard yet still reproduce a cherry-picked result","An effect-size or materiality rule, which asks whether a decline matters but not how many opportunities existed to find it","Representative-use measurement, which addresses biased circulation evidence rather than false discoveries created by repeated slicing"],"negative_tests":{"intervention_falsifier":"In a shadow comparison, the register and staged rule are falsified as an operative guardrail if undocumented looks still enter reports, exploratory leads reach the action queue, or confirmation results can be reproduced only after changing the locked claim specification.","problem_falsifier":"The proposed problem is absent if each consequential collection claim is genuinely single, pre-specified before data access, evaluated through one fixed analysis path, and retained with all outcomes, with no dashboard or analyst search generating alternative opportunities.","risks":["A strict family definition may suppress useful leads when inexpensive follow-up would be acceptable.","Staff may evade the register through offline spreadsheets or informal dashboard browsing.","Dependence among circulation comparisons may make the chosen adjustment unnecessarily conservative.","A subsequent period may differ because of seasonality, closures, catalog changes, or programming rather than confirmation failure.","Status labels may become ceremonial while decision narratives still overstate exploratory results.","Aggregate circulation omits in-library use, preservation value, curricular relevance, and community importance.","Small-cell reporting could expose sensitive borrowing patterns unless suppression rules are enforced."],"strongest_counterevidence":"If reconstructed review histories show that highlighted low-use patterns were already specified before inspection, all alternative slices were retained, and fresh-period confirmation routinely preceded collection action, hidden multiplicity would not explain the observed decision process."},"next_evidence_step":"Run one read-only shadow pilot on the next scheduled review for at most two branches and ten predeclared classification bands. Before opening results, freeze the family definition, eligible filters, metrics, windows, Holm rule, privacy suppression, and confirmation specification. Reconstruct and register every automated and manual look, generate status-labeled reports, and test whether any lead can reach the action queue without the required record. Reserve the following review period for confirmation; make no collection changes during the pilot. End after one discovery cycle and one confirmation check, reporting only process adherence, lead-status transitions, missing looks, and contamination incidents.","observable_state":"A quarterly dashboard permits repeated slicing of the same circulation and holdings records by branch, classification band, metric, eligibility filter, and time window; the final weeding packet shows a selected decline but does not expose a machine-readable count and specification of all inspected alternatives.","prior_art_status":"UNSEARCHED","problem":"During collection review, library staff can search many combinations of branches, classification ranges, circulation measures, item filters, and time windows, then present the most pronounced apparent decline as evidence that a subject area should be weeded or deprioritized. Because the unselected looks are not carried into the review packet, the chosen decline is interpreted as if it were the sole planned claim.","proposal_index":1,"remaining_contrastive_claim":"The intervention addresses credibility lost specifically because a consequential collection claim was selected from many attempted looks: unlike a fixed usage threshold, rerunnable workflow, or materiality check, it makes the full opportunity set determine claim status and requires untouched evidence before action eligibility.","revision_record":{"claim_changes":[],"conceptual_changes":[],"evidence_changes":[],"operational_changes":[],"parent_version":null,"progress_targets_addressed":["Instantiate the archetype in a concrete library collection-management problem","Preserve the many-looks-to-false-positive causal structure","Define bounded authority, privacy safeguards, falsifiers, risks, and a non-action shadow pilot"]},"schema_version":1,"structural_mapping":[{"archetype_element":"Many attempted claims","domain_realization":"Dashboard slices across branches, classification bands, circulation metrics, item filters, and time windows within one collection-review cycle"},{"archetype_element":"Ordinary single-claim interpretation","domain_realization":"The selected low-use slice is presented under an unadjusted threshold without its alternative slices."},{"archetype_element":"Inflated false-positive opportunity","domain_realization":"Repeated slicing creates multiple chances for ordinary variation to resemble a meaningful local-use decline."},{"archetype_element":"Claim family","domain_realization":"The frozen set of all permitted branch-by-classification-by-metric-by-filter-by-window comparisons for the review cycle"},{"archetype_element":"Multiplicity inventory and discovery record","domain_realization":"An append-only register of queries, filters, specifications, null results, selected leads, owners, and status changes"},{"archetype_element":"Error-risk policy and adjusted rule","domain_realization":"Familywise control for leads that could support withdrawal review, with other patterns retained only as exploratory"},{"archetype_element":"Exploratory–confirmatory boundary","domain_realization":"Discovery-period circulation identifies leads; an untouched subsequent period tests a locked claim."},{"archetype_element":"Claim-status labeling","domain_realization":"Exploratory, adjusted-discovery, confirmed, rejected, and decision-eligible labels displayed in reports and workflow queues"},{"archetype_element":"Confirmation before costly action","domain_realization":"No transfer, storage, withdrawal, or acquisition change is eligible until fresh-period confirmation and branch-context review are complete."}],"title":"Multiplicity-Aware Gate for Collection-Weeding Claims","version":0}