{"schema_version":1,"experiment_id":"eoa_inverse_innovation_exp06_four_proposal_generalization60_20260803","source_assessment_id":"bounded_rivalry_governance__criminology_forensic:P3:v0","cell_id":"bounded_rivalry_governance__criminology_forensic","proposal_index":3,"qualification":"EMPIRICAL_PARTNER_CANDIDATE","criteria":{"specific_differentiated_claim":{"status":"YES","reason":"The proposal retains a specific incremental claim: compared with conventional case conferences and ordinary ACH, separated teams using identical cutoff records, preregistered predictions, equal budgets, rights-adjusted scoring, and limited validation slots will improve held-out calibration and information value without increasing specified harms."},"credible_problem_signal":{"status":"YES","reason":"Official guidance and primary research support scarce cold-case capacity, inconsistent use of impartial investigative reasoning, and a credible tunnel-vision risk, while appropriately leaving the exact prevalence of incumbent-control and rivalry mechanisms unresolved."},"identifiable_partner_or_adopter":{"status":"YES","reason":"Cold-case or major-case review units are concrete adopters and authorizers; an agency with secure closed-case records, qualified investigators, and research authority can host the proposed retrospective study."},"partner_access_is_necessary":{"status":"YES","reason":"The decisive uncertainties concern comparative performance, panel reliability, leakage, gaming, entrenchment, and workflow behavior using realistic time-split records and qualified participants. These require proprietary records, secure infrastructure, and practitioner participation and cannot be resolved through public web research."},"safe_authorized_first_step":{"status":"YES","reason":"The first step is a retrospective, masked, no-case-impact study using copied closed-case or synthetic records, with no witness contact, original-evidence use, coercive action, live resource allocation, or operational interpretation. Explicit halt, voiding, access-revocation, and ranking-suppression rules bound failures."},"bounded_decisive_empirical_design":{"status":"YES","reason":"The preregistered 8–12-packet crossover compares the gate against both conventional conferences and noncompetitive ACH under equal records and analyst time, uses blinded held-out scoring and prespecified harm measures, and has explicit success, falsification, exclusion, and halt rules."},"no_material_negative_gate":{"status":"YES","reason":"All substantive pipeline gates are YES. Cost scope is uncertain, but that uncertainty is not a safety, authority, problem, adopter, claim, or test-design failure and can be scoped alongside the partner-specific study."},"not_merely_more_research":{"status":"YES","reason":"The next action specifies the partner, records, three conditions, sample range, crossover structure, controls, captured variables, outcomes, harm thresholds, leakage rules, and advancement decision rather than calling generically for more research."}},"uncertainty_types":["PROBLEM_PREVALENCE","ADOPTER_PULL","INCREMENTAL_EFFECT","WORKFLOW_FIT","DATA_ACCESS","COST_SCOPE"],"partner_profile":"An authorized cold-case or major-case review unit with lawful access to 8–12 sufficiently complete closed, maskable case records; a secure research environment; qualified investigators or analysts; independent reviewers; and authority to approve a retrospective study with no operational consequences.","required_access":"Lawfully prepared, citation-addressable, masked time-split record copies; qualified participant time; secure role-based access and logs; independent scoring capacity; packet-completeness and recognition-risk assessment; and written local approvals covering research authority, privacy, disclosure, retention, labor protections, and prohibition on operational use.","bounded_empirical_test":"Preregister a three-condition retrospective crossover across 8–12 masked, time-split closed-case or synthetic packets: the bounded hypothesis gate, a conventional multidisciplinary case conference, and a noncompetitive ACH worksheet. Equalize cutoff records and analyst hours, rotate qualified participants, blind reviewers to condition, reveal later segments only after predictions are locked, and measure calibration, discriminating information value, citations, unsupported allegations, redundancy, panel reliability, time, privacy burden, evidence consumption, strategic withholding, and entrenchment.","success_condition":"The gate outperforms the better comparator on preregistered held-out prediction calibration and rights-adjusted information value, meets the minimum panel-reliability threshold, and does not cross any prespecified threshold for unsupported allegations, strategic withholding, entrenchment, delayed sharing, privacy burden, evidence consumption, or scorer inconsistency.","falsification_condition":"Reject the incremental claim if the gate fails to outperform the better comparator on either primary outcome, worsens confidence calibration, falls below the panel-reliability threshold, or increases any prespecified harm. Void affected packets if masking fails, a participant recognizes a case, records are unequal, or later-outcome information leaks to analysts or scorers.","rationale":"This is a genuine empirical-partner case despite substantial adjacent prior art. Public evidence establishes a credible problem, organizational hosts, a differentiated comparison, and a safe test, but cannot answer the decisive questions about incremental effect and behavior in realistic secure workflows. A partner is therefore necessary, and the proposed retrospective crossover is sufficiently bounded and falsifiable to determine whether further evaluation should stop or advance."}