Use transparent evidence, explicit constraints and reproducible calculations to examine civilian policy. This pathway changes the questions and tool selection; it does not change the mathematics or grant decision authority.
public tenders, awards, amendments, disclosed authorizations and amounts; CORE-017; output missing records and documented inconsistencies, not crime accusations.
Shared calculator exercise.
Try Human Revision and Decision Recordpublic allegations or policy justifications, original source records and independent confirmations; CORE-005; output evidence relationships and contested status.
Shared calculator exercise.
Try Evidence provenance graphdated public complaints, acknowledgments, review disposition and correction dates; compute elapsed time, known/unknown cases; output transparent process latency.
Planned.
Read the proposed methodproject IDs, award IDs, periods, recipients and declared scope; flag potential overlap with human-review labels; never presume fraud from duplicate names or addresses.
Planned.
Read the proposed methodproposed public administrative process with required notice, appeal, decision ownership and retention; compare with specified framework and legal citations; output missing steps for counsel review, not legal compliance certificate.
Planned.
Read the proposed methodauditor/decision-maker affiliations openly declared for a proceeding, origin of reports, review roles; graph declared conflicts and independence gaps, do not infer wrongdoing or undisclosed relationships.
Planned.
Read the proposed methodState the question, affected civilian service, geography and period. Separate dated observations from assumptions. Identify the current baseline and a structured non-AI alternative. Record resources, lawful authority, burdens, fallback and the evidence that would change the decision.
Report unknown costs as unknown. Preserve conflicting claims with their original sources. Obtain the relevant human review before using a calculation to support a public decision. Campaign and candidate research is limited to public policy analysis, not voter profiling or personalized persuasion.
Prepare an executive briefExamine the original MAR assumptionsHow many declared traceable origin clusters support these reports, and which origins remain unknown?
Unit: records. Validation: computational unit tests; no empirical calibration.
Open calculator & methodWhat changed, who recorded the decision, why, and which provenance or review records are missing?
Unit: records. Validation: computational unit tests; no empirical calibration.
Open calculator & methodNo. The same inputs and methods produce the same results for every user. No endorsement, voter profile or election prediction is generated.
Keep missing values unknown, disclose the gap and use the result only within its stated scope. A planned module is not an operational prediction.