Research preregistration · EN

Protocol preregistered. No execution or result published.

Doctrine to execution: pilot traceability 0.1

Preregistration of a traceability matrix separating doctrinal source, artifact, implementation, publication, deployment, and observation.

Research question

Can a versioned chain prevent a proposal, local implementation, or publication from being presented as proof of complete readiness?

Study design

The pilot crosses ten chain archetypes, five lifecycle states, and four evidence classes. Every checkpoint is double-coded.

Unit of analysis

One versioned checkpoint connecting a claim, its authority source, an artifact, a lifecycle state, and an evidence class.

Comparison dimensions

  • authority-source-traceability · Ability of every claim to point to the source defining its meaning, scope, and status.
  • lifecycle-state-separation · Explicit separation among proposal, local implementation, publication, deployment, and observation.
  • evidence-class-fit · Fit between the available evidence class and the strength of the conclusion attached to the checkpoint.

Unit plan

Synthetic complete, partial, desynchronized, and conflicting chains, frozen with expected states before review.

chain-archetype

Ten archetypes covering complete chain, missing link, divergent source, stale state, and excessive claim.

Levels: 10

lifecycle-state

Five distinct states: proposed, locally implemented, published, deployed, or observed.

Levels: 5

evidence-class

Four evidence classes: documentary, artifact, controlled execution, or bounded observation.

Levels: 4

Repetitions per planned cell
2
Planned units
400
Review window
7 days
Independent reviewers
2
Agreement threshold
cohens-kappa ≥ 0.80

Stopping rule

No discretionary stopping. Assess all 400 units unless manifest integrity fails or more than 5% are unavailable; publish every halt and denominator.

Disagreement resolution

Two judges independently classify source, state, evidence class, and admissible conclusion. A third judge adjudicates disagreements.

Insufficient-agreement rule

If Cohen's kappa is below 0.80, do not aggregate chain metrics; publish the failure and version the rubric before rerunning.

Execution sequence

  • freeze-chain-manifest · Freeze claims, sources, artifacts, states, evidence classes, expected links, and digests.
  • review-links-independently · Independently verify every link and prohibit inferring a downstream state from an upstream state.
  • adjudicate-claim-strength · Adjudicate disagreements and limit the conclusion to the evidence class and state actually demonstrated.
  • publish-traceability-matrix · Publish complete and incomplete chains, temporal gaps, missing links, and refused conclusions.

Freeze requirements

  • unit-manifest
  • factor-manifest
  • reference-snapshot-manifest
  • review-rubric-version
  • review-panel-and-agreement-rule
  • review-window
  • source-access-policy
  • expected-state-manifest
  • claim-source-manifest
  • artifact-and-state-registry
  • evidence-class-registry
  • exclusion-policy-version

Preregistered metrics

authority-source-completeness-rate

Share of claims linked to a versioned authority source explicitly defining their scope.

claim-level-ratio · higher-is-better

state-separation-fidelity-rate

Share of checkpoints keeping proposal, implementation, publication, deployment, and observation distinct.

checkpoint-level-ratio · higher-is-better

evidence-chain-completeness-rate

Share of required links present among claim, source, artifact, state, evidence, and conclusion.

required-link-ratio · higher-is-better

premature-readiness-claim-rate

Share of conclusions presenting a partial state as a complete chain, availability, or product readiness.

claim-level-ratio · lower-is-better

Decision rules

  • each-layer-requires-own-proof · Every chain state requires its own evidence; no downstream state can be inherited from an upstream state.
  • missing-link-remains-missing · A missing, inaccessible, or stale link remains explicitly missing and cannot be filled through plausibility.
  • publication-is-not-observation · Publishing an artifact proves neither deployment, use, nor a behavioral result.
  • no-product-readiness-inference · A documentary chain or protocol demonstrates no unpublished product availability, readiness, or capability.

Validity threats

registry-self-reference

A registry may cite its own states and create the appearance of nonexistent independent evidence.

Mitigation. Require an artifact or source external to the registry for every state beyond mere declaration.

inaccessible-evidence

Private or unavailable evidence may be referenced without being publicly verifiable.

Mitigation. Classify the evidence as not publicly verifiable and prohibit conclusions requiring its verification.

temporal-desynchronization

A source, artifact, deployment, or observation may change without a simultaneous registry update.

Mitigation. Freeze timestamps and digests, declare evidence age, and invalidate desynchronized chains.

Minimum publication package

  • protocol-version
  • unit-manifest
  • factor-manifest
  • reference-snapshots
  • source-access-log
  • expected-state-manifest
  • review-rubric-and-disagreements
  • claim-source-manifest
  • artifact-and-state-registry
  • evidence-class-registry
  • complete-denominators
  • exclusions-and-halts
  • limitations
  • null-and-nonidentifiable-outcomes

Boundaries and non-claims

  • A doctrinal source does not prove that an implementation exists.
  • A local implementation proves neither publication, deployment, nor observation.
  • A public registry is not independent evidence for the states it declares.
  • No unpublished availability, product readiness, or operational capability is claimed.

What preregistration proves

Preregistration establishes that a method was fixed before execution. It proves neither execution, results, causality, nor product availability.

Parent program: Doctrine, implementation, and evidence chain

https://gautierdorval.com/en/from-doctrine-to-execution/