{"actors":["Municipal inspector-general intake analyst","Residents who spoke or petitioned about a municipal policy","Municipal code-enforcement unit and inspection supervisors","Enforcement-data steward","Inspector-general counsel or designated procedural reviewer"],"affected_objective":"Impartial, reviewable oversight of municipal enforcement allocation while protecting residents from retaliation and employees from unsupported accusation.","arm":"COMMON_P1","authority_safety":{"authorized_first_step":"The intake analyst may conduct a read-only comparison of already authorized complaint timestamps, routing metadata, inspection assignments, geographic or hazard factors, and source-system provenance, then record a provisional ranking and the least-intrusive next probe.","decision_authority":"The memo author may prioritize reversible fact-finding and recommend preservation of records; only officials holding existing investigative, adjudicative, or disciplinary authority may compel evidence, make legal findings, impose remedies, or sanction personnel.","excluded_actions":["Treating the working explanation as a finding of retaliation or misconduct","Naming suspected employees to the public or unrelated personnel","Changing, suspending, or disciplining staff solely from the explanatory ranking","Using protected characteristics, political viewpoint, or legally protected activity as an enforcement-risk score","Accessing records outside existing authorization or bypassing privilege, confidentiality, retention, or notice rules","Allowing the memo to substitute for applicable burdens of proof, hearing rights, or independent adjudication"],"halt_rollback":"Halt use if the observation record cannot be separated from allegations, required comparison data were unlawfully obtained, protected groups receive asymmetric scrutiny, or reviewers repeatedly treat rankings as findings. Withdraw the memo from operational routing, preserve an access-controlled audit copy, notify the designated reviewer, and return the matter to ordinary intake pending correction."},"baseline":"Ordinary complaint-by-complaint inspector-general intake in which an analyst reads narratives, checks selected records, and exercises unstructured discretion about whether retaliation, routine enforcement, or error deserves follow-up; rival explanations and revision triggers may remain implicit.","candidate_id":"abductive_explanation_selection__law_governance__COMMON_P1","causal_chain":["A cluster of inspections following residents' protected participation creates a trace compatible with retaliation, routine complaint-driven enforcement, correlated property conditions, scheduling coincidence, or data error.","Unstructured intake permits the most vivid narrative—either retaliation or routine enforcement—to shape which records are sought.","The intervention freezes a neutral observation record before interpretations are entered.","A bounded candidate set forces inclusion of the leading account, base-rate-common accounts, a high-impact retaliation account, and an action-changing data-integrity rival.","An explicit fit matrix compares timing, coverage, mechanism, background context, missing evidence, anomalies, and predicted traces without converting plausibility into a legal finding.","If one explanation materially fits better, it becomes an action-limited working explanation; otherwise the memo preserves a ranked tie.","A probe chosen for its ability to distinguish the leading candidates directs limited fact-finding toward complaint origin, queue departures, assignment provenance, or source-system reconciliation.","Failed predictions, new records, or a stronger rival reopen the ranking before any formal investigative or adjudicative conclusion."],"cell_id":"abductive_explanation_selection__law_governance","consequence":"Without a reviewable explanatory frame, oversight may prematurely accuse staff, dismiss a retaliation signal as coincidence, collect records that only confirm the intake analyst's first story, or escalate an ambiguous pattern into intrusive investigation without a clear evidentiary purpose.","diversity_from_prior_proposals":"No comparison with prior proposals was performed under runtime isolation. Within the supplied domain, this candidate focuses specifically on pre-adjudicative explanation selection for an enforcement-allocation anomaly, with authority limited to reversible oversight triage.","experiment_id":"eoa_inverse_innovation_exp13_second_slot_policy60_20260806","intervention":"Introduce an access-controlled Retaliation-Signal Explanation Docket for municipal inspector-general intake. For each qualifying inspection cluster, the analyst records the direct trace separately from allegations; bounds the event, population, and time window; lists routine complaint routing, correlated code conditions, scheduling coincidence, data-linkage error, retaliatory targeting, and any case-specific rival; and compares them in a standard matrix covering evidentiary coverage, timing, mechanism, base-rate context, anomalies, missing evidence, and discriminating predictions. The docket selects a provisional, action-limited leader only when it materially outperforms the nearest rival; otherwise it records an unresolved ranked set. It then authorizes only the least-intrusive evidence check capable of separating the live accounts and states prediction failures that require revision or abandonment.","mechanism_mapping":[{"counterfactual_removal":"Without the matrix, candidate comparison returns to implicit analyst preference and the reason one account outranks another is not reviewable.","mechanism_slug":"inference_to_best_explanation_matrix","role":"Compares each candidate against the same evidence, timing, mechanism, missing-evidence, and anomaly criteria."},{"counterfactual_removal":"Without the case memo, observations, allegations, confidence limits, and action limits can blur during handoff.","mechanism_slug":"explanatory_case_memo","role":"Preserves the explanandum, rival ranking, provisional status, ownership, and authorized use in one controlled record."},{"counterfactual_removal":"Without a disconfirming probe, fact collection can follow the favored narrative and leave the nearest rival untouched.","mechanism_slug":"disconfirming_probe_plan","role":"Chooses the smallest authorized record check whose possible results differ across the live explanations."},{"counterfactual_removal":"Without the log, later evidence may be absorbed into the original story without showing when or why the explanation changed.","mechanism_slug":"abduction_log","role":"Records predictions, incoming evidence, rank changes, tie states, and revision decisions over time."}],"nearest_rivals":["A disparity dashboard that flags temporal or geographic inspection concentration but does not compare explanations or specify defeaters","A formal retaliation investigation that invokes compulsory or adjudicative procedures before intake ambiguity has been narrowed","A statistical causal audit designed to estimate whether protected participation changed inspection likelihood","Random audit sampling that checks enforcement files without using the anomaly to select discriminating records"],"negative_tests":{"intervention_falsifier":"In a bounded shadow review, the structured docket fails to produce auditable observation-versus-interpretation separation, repeatedly yields probes that do not distinguish its named rivals, or causes reviewers to anchor on retaliation more strongly than the existing intake record despite action-limit labels.","problem_falsifier":"Existing objective routing rules and complete provenance records already determine why the inspections occurred, or the available evidence already satisfies the applicable formal proof process, leaving no genuine hidden-cause ambiguity for provisional explanation selection.","risks":["The retaliation candidate may become an accusation merely by being written down.","Standard candidate lists may omit locally specific mechanisms or manufacture false balance.","Analysts may encode political, geographic, or demographic bias in the explanandum boundary or fit criteria.","Sensitive speech, complaint, property, or personnel data may be exposed through the docket.","A provisional memo may leak into discovery, personnel files, public-record responses, or adjudication without its limits.","The added structure may delay urgent record preservation or protection when an independently authorized emergency threshold is met.","A tidy single-cause account may obscure compound causes such as routine complaints amplified by selective supervision."],"strongest_counterevidence":"Authenticated records show that independent complaints and rule-based queue placement predated the residents' protected activity, that the same assignment rules were applied without unexplained departures, and that source-system reconciliation accounts for the apparent temporal cluster."},"next_evidence_step":"Conduct a read-only shadow review of no more than 12 closed municipal oversight matters involving alleged selective or retaliatory inspection. Two authorized reviewers use only the records available at original intake: one reconstructs the ordinary intake rationale and the other completes the docket without affecting any person or case. Check whether the docket keeps observation separate from allegation, includes an action-changing rival, identifies a genuinely discriminating authorized probe, preserves ties when warranted, and states a usable revision trigger; stop after the sample and submit only de-identified process findings to the designated procedural reviewer.","observable_state":"An inspector-general intake packet shows that residents who recently spoke, petitioned, or complained about a municipal policy were subsequently selected for code inspections in a temporal or administrative cluster. Directly observable items are participation dates, complaint receipts, inspection creation and assignment timestamps, stated inspection grounds, locations, queue positions, routing fields, and data provenance. Retaliation, routine complaint response, correlated code conditions, scheduling coincidence, supervisor intervention, and linkage error are interpretations to be compared, not observations.","prior_art_status":"UNSEARCHED","problem":"When municipal inspections cluster after residents engage in protected civic activity, an oversight office must decide what limited fact-finding is justified before it knows whether the pattern reflects retaliatory targeting, routine enforcement inputs, correlated conditions, scheduling coincidence, or a records artifact. Existing intake discretion can let either suspicion or institutional reassurance harden into a factual conclusion before credible rivals and disconfirming traces are recorded.","proposal_index":1,"remaining_contrastive_claim":"Relative to anomaly flagging, immediate formal investigation, causal estimation, or random audit, the docket's distinct claim is limited to the reasoning transition between a puzzling enforcement trace and a reversible next inquiry: it makes rival explanatory fit, unresolved ties, defeaters, and action limits explicit without supplying a legal finding or effect estimate.","revision_record":{"claim_changes":["Initial version; no prior claim was revised."],"conceptual_changes":["Initial version maps abductive explanation selection to pre-adjudicative municipal enforcement oversight."],"evidence_changes":["No external evidence was searched or added."],"operational_changes":["Initial version bounds use to read-only shadow review and reversible investigative triage."],"parent_version":null,"progress_targets_addressed":["Concrete explanandum and actors specified","Observation separated from interpretation","Base-rate, high-impact, and action-changing rivals retained","Defeasibility, authority limits, falsifiers, risks, and revision triggers made explicit","First evidence step bounded and non-operational"]},"schema_version":1,"structural_mapping":[{"archetype_element":"Surprising Observation Record","domain_realization":"Timestamped civic participation and subsequent inspection-routing traces are recorded without labeling the sequence retaliatory."},{"archetype_element":"Explanandum Boundary","domain_realization":"The docket fixes the residents or properties, protected activities, inspection events, administrative units, and time window being explained."},{"archetype_element":"Candidate Explanation Set","domain_realization":"The set includes ordinary complaint routing, correlated code conditions, scheduling coincidence, data error, retaliatory targeting, and a case-specific rival that would redirect action."},{"archetype_element":"Explanatory Fit Criteria","domain_realization":"Candidates are compared on trace coverage, temporal order, routing mechanism, background enforcement context, anomalies, unsupported assumptions, missing records, and divergent predictions."},{"archetype_element":"Rival Comparison Record","domain_realization":"The memo states why the leader currently exceeds its nearest rival or why their fit remains tied."},{"archetype_element":"Best-So-Far Explanation","domain_realization":"The leading account guides only record preservation and limited fact-finding, not liability, discipline, or public attribution."},{"archetype_element":"Defeasibility Status Label","domain_realization":"Each output is labeled working and action-limited, contested, or unresolved tie, with its evidentiary limits."},{"archetype_element":"Discriminating Evidence Plan","domain_realization":"The docket identifies a minimal authorized check of complaint origins, queue departures, assignment metadata, contemporaneous instructions, or data reconciliation whose outcomes separate live accounts."},{"archetype_element":"Action–Confidence Boundary","domain_realization":"Provisional fit may justify read-only review or preservation, while compulsory process, accusation, remedy, and sanction remain subject to existing authority and proof standards."},{"archetype_element":"Revision Trigger","domain_realization":"The ranking reopens when predicted routing traces fail, provenance changes, new independent complaints emerge, a rival gains coverage, or the leader requires ad hoc exceptions."},{"archetype_element":"Explanation Owner","domain_realization":"A named intake analyst owns updates, while a designated procedural reviewer checks boundaries, access, and escalation."}],"title":"Action-Limited Retaliation-Signal Docket for Municipal Inspection Oversight","version":0}