{"schema_version":1,"experiment_id":"eoa_inverse_innovation_exp09_archetype_breadth150_20260804","cell_id":"search_space_pruning__accounting_auditing","arm":"BREADTH_PROBE_ONE_SHOT","candidate_id":"search_space_pruning__accounting_auditing__P1","proposal_index":1,"version":0,"title":"Auditable Pruning of Intercompany Reconciliation Investigation Branches","problem":"When a consolidated close contains an unexplained intercompany elimination difference, accountants may face a combinatorial set of possible entity-pair, account, document, currency, timing, and transformation paths. Fully tracing every possible path is impractical, while informal narrowing can discard the branch containing the actual accounting error without leaving a reviewable reason.","actors":["Group controller","Consolidation accountant","Entity accountants","Internal audit reviewer"],"observable_state":"A material or control-significant intercompany elimination difference remains unresolved, and the reconciliation workspace contains many candidate mismatch branches linking ledger entries across entity pairs, accounts, currencies, periods, documents, and consolidation adjustments. Investigators repeatedly inspect branches that contradict known posting facts, while other branches disappear through undocumented judgment.","consequence":"Close effort is consumed by predictably infeasible investigations, resolution may be delayed, and an accounting error can remain uncorrected or be cleared through an unsupported adjustment because the true branch was prematurely excluded.","affected_objective":"Identify and substantiate the transaction or transformation path responsible for the elimination difference while preserving an auditable basis for exclusions and keeping viable or uncertain causes reachable.","intervention":"Represent each possible explanation as a branch from the observed difference through entity pair, reciprocal account, source document, posting period, transaction currency, exchange-rate treatment, and consolidation adjustment. Apply explicit constraint filtering only where ledger facts make a branch impossible; defer rather than exclude branches with missing evidence. Remove dominated branches when they require all the same unresolved assumptions as another branch plus additional contradictions. Retain representatives from unusual transaction classes, record every exclusion, sample excluded branches for false negatives, and allow reentry when documents, mappings, periods, or consolidation rules change.","structural_mapping":[{"archetype_element":"Search Space Definition","domain_realization":"All ledger-to-consolidation explanation paths capable of contributing to the specified intercompany elimination difference."},{"archetype_element":"Candidate or Region Representation","domain_realization":"A branch identified by entity pair, reciprocal accounts, source-document linkage, posting period, currency treatment, and consolidation-adjustment sequence."},{"archetype_element":"Objective or Success Criterion","domain_realization":"A branch remains viable if it can reconcile to the observed difference and support a documented accounting explanation or correction."},{"archetype_element":"Constraint Filter","domain_realization":"Exclude a branch only when authoritative ledger or consolidation facts contradict mandatory attributes such as entity identity, account eligibility, period availability, currency path, document linkage, or adjustment scope."},{"archetype_element":"Feasibility Test","domain_realization":"Recompute whether the branch could contribute to the difference after sign, amount, currency, timing, and consolidation transformations."},{"archetype_element":"Dominance Check","domain_realization":"Remove a branch that explains no additional amount or evidence and contains every contradiction of a retained branch plus further unsupported assumptions."},{"archetype_element":"Diversity Preservation Rule","domain_realization":"Keep at least one unresolved representative from each implicated transaction class and transformation type rather than retaining only familiar invoice-matching explanations."},{"archetype_element":"False-Negative Review","domain_realization":"A reviewer examines a bounded sample of excluded branches, including unusual and high-value items, to test whether exclusion rules removed plausible causes."},{"archetype_element":"Reentry or Exception Path","domain_realization":"An excluded branch returns to the active set when new source documents, corrected mappings, reopened periods, or changed consolidation rules invalidate its exclusion premise."},{"archetype_element":"Pruning Audit Trail","domain_realization":"For every excluded branch, the workspace records the rule, supporting fields, assumptions, timestamp, rule version, and responsible reviewer."},{"archetype_element":"Pruning Owner","domain_realization":"The group controller approves pruning criteria; the consolidation accountant applies them; internal audit may challenge exclusions but does not post corrections."}],"mechanism_mapping":[{"mechanism_slug":"constraint_filtering","role":"Eliminates investigation branches contradicted by authoritative entity, account, period, document, currency, or consolidation facts before costly document tracing.","counterfactual_removal":"Without constraint filtering, investigators must continue tracing branches already known to be infeasible, so the search remains overloaded."},{"mechanism_slug":"dominated_option_removal","role":"Removes explanations that cannot account for more of the difference than a retained branch and require additional contradicted assumptions.","counterfactual_removal":"Without dominance checks, redundant inferior explanations remain active and consume the same scarce review capacity as distinct viable causes."},{"mechanism_slug":"sample_audit_of_exclusions","role":"Tests a stratified sample of pruned branches for plausible missed causes and triggers rule review when a false exclusion is found.","counterfactual_removal":"Without exclusion sampling, the process can appear tractable while systematically hiding false negatives produced by an incorrect or stale rule."}],"causal_chain":["An elimination difference generates many possible ledger-to-consolidation explanation branches.","Explicit representation makes the candidate branches and their required assumptions visible.","Hard ledger and consolidation constraints remove branches that cannot produce the observed difference.","Dominance checks remove branches that add contradictions without adding explanatory reach.","Diversity holdouts keep unusual transaction and transformation families available where evidence is incomplete.","The remaining branch set becomes small enough for document tracing and substantive accounting review.","Exclusion logs, sampled false-negative review, and reentry triggers expose or reverse premature pruning.","Investigators can pursue a supportable cause or correction without treating undocumented narrowing as evidence."],"baseline":"Investigators query ledgers, request documents, and follow likely entity or account paths using personal experience and ad hoc spreadsheets. Narrowing occurs through undocumented judgments, infeasible branches may be revisited by different staff, and discarded branches lack consistent reentry or false-negative review.","nearest_rivals":["Risk-based transaction sampling, which selects items for audit testing but does not structure and prune competing causal paths for a known reconciliation difference.","Automated account matching, which pairs reciprocal entries but does not govern elimination of broader timing, mapping, currency, or consolidation-adjustment explanations.","Materiality thresholds, which determine the significance or required treatment of a difference rather than which causal branches remain feasible.","Root-cause analysis checklists, which enumerate common causes but do not explicitly eliminate infeasible regions with logged criteria and reentry safeguards."],"remaining_contrastive_claim":"The candidate's distinguishing claim is limited to process structure: for a known intercompany difference, it governs removal of impossible or dominated causal investigation branches while retaining uncertain branch families and auditable reentry, rather than merely sampling transactions, matching entries, applying materiality, or listing common causes.","authority_safety":{"decision_authority":"The group controller owns and approves pruning criteria and any accounting disposition; authorized accountants investigate retained branches. The intervention itself has no authority to post, reverse, waive, or certify entries.","authorized_first_step":"Run a read-only retrospective replay on one previously completed reconciliation using frozen ledger extracts and compare proposed exclusions with the documented final explanation.","excluded_actions":["Posting or reversing journal entries","Changing consolidation mappings or exchange-rate rules","Suppressing unresolved differences from close reporting","Treating a pruned branch as proof that no error exists","Excluding branches solely because they are inconvenient, unfamiliar, immaterial without approved policy, or missing evidence","Replacing required audit procedures, professional judgment, or approval controls"],"halt_rollback":"Stop the replay if a hard filter excludes the documented cause, authoritative fields cannot support the rule, or exclusion records cannot be reconstructed. Restore all branches to the candidate set, invalidate the implicated rule version, and require controller review before any further use."},"negative_tests":{"strongest_counterevidence":"The strongest counterevidence would be a retrospective set in which the documented causes are frequently located in branches that the proposed hard rules exclude, especially because source-system identifiers, timing fields, mappings, or currency transformations are unreliable.","problem_falsifier":"The problem framing is falsified if completed reconciliations show that candidate paths are few enough for complete review, evaluation cost is not constraining, and undocumented or repeated narrowing is absent.","intervention_falsifier":"The intervention is falsified if, under read-only replay, it excludes the documented cause, fails to reduce branches requiring substantive review, or creates more rule validation and exclusion-review work than direct examination of the full candidate set.","risks":["Incorrect source-system fields may make a feasible branch appear impossible.","Preferences such as familiarity or low document-retrieval cost may be mislabeled as hard constraints.","Coarse rules may remove rare mapping, timing, foreign-exchange, or manual-adjustment causes.","A retained-class quota may provide token diversity while missing variation within a class.","Users may mistake exclusion for evidence that the remaining branch is correct.","Rules may become stale after chart-of-accounts, entity, system, or consolidation-policy changes.","Detailed pruning criteria may be gamed to keep problematic transactions outside active review."]},"next_evidence_step":"Using frozen, read-only data from one completed close and one resolved intercompany difference, enumerate the branches that were available before resolution. Apply no more than five controller-specified hard constraints, log every proposed exclusion, retain one representative per implicated transaction or transformation class, and have an accountant blinded to the rules inspect all retained branches plus a stratified sample of up to 20 excluded branches. Record whether the documented cause remained reachable, which exclusions were unsupported, and how many branches still required substantive review; do not operationalize the rules from this replay alone.","prior_art_status":"UNSEARCHED","diversity_from_prior_proposals":"Not assessed because this is an isolated one-shot proposal and no other experiment candidates were inspected.","revision_record":{"parent_version":null,"progress_targets_addressed":[],"conceptual_changes":[],"operational_changes":[],"evidence_changes":[],"claim_changes":[]}}