The site uses explicit claim states, artifact checksums, negative-result reporting, and fail-closed release gates. These rules apply to formal claims, computational results, standards proposals, and numerical measurements.
Canonical public record: https://decisionaccounting.org/methods/
STATUS VOCABULARY
One meaning for each public evidence state.
State
Public meaning
Proved / verified
Reserved for a formal claim whose exact current hash has the required proof and same-hash independent verification evidence. The current Laboratory records zero verified theorems.
Empirically supported
A stated result tied to an inspectable design, data or saved simulation cells, estimand, uncertainty treatment, and reproduction path.
Calibrated
A measurement whose sources, boundaries, denominators, assumptions, uncertainty, and result receipt have passed the public admission gate.
Provisional
A claim with a formal statement or argument that still has proof gaps, unresolved objections, or incomplete independent review.
Proposed
A standard, policy, mechanism, or implementation that is offered for testing and has not been established by adoption or field validation.
Disputed
A claim with a material objection or counterexample that remains unresolved in its public record.
Held
A public release gate is closed. The claim or value must not be used as an admitted result.
Inspectable design
Computational experiments
A result must identify the environment, treatment, comparator, seeds or sampling, estimand, saved outputs, uncertainty procedure, code path, and known implementation limits. Nulls and trigger failures remain in the public result.
Plain-English claims precede formalism. Assumptions, formal statement, proof or proof gaps, evidence, objections, boundaries, counterexamples, reforms, relationships, lineage, and falsification conditions belong to the same versioned record. Verification requires the final same-hash evidence chain.
Each numerical release needs an independently sourced denominator, same-boundary welfare channels, declared valuation rules, distribution and dependence assumptions, reproducible code, an exact result hash, row admission, rank policy, and portfolio tie-out. The current public numerical gate is closed.
Material corrections identify the affected artifact, date, version, previous public statement, current statement, and successor. Frozen downloads remain addressable, while the current page links to the controlling version.
A correctly implemented record with active authority and consequence could fail to improve reconstruction or decision quality in a well-powered field test.
Existing governance or provenance systems could already provide the same pre-decisional evidence, authority, rejected-alternative, prediction, system-effect, and supersession functions at lower cost.
A formal claim can fail through an invalid derivation, hidden assumption, valid counterexample, scope error, or mismatch between its prose and formal statement.
A measurement can fail when source boundaries, denominators, overlap rules, valuations, uncertainty, or reproducibility do not survive independent audit.