ORHSA: Claim-State Quantum for evidence governance

[neuron7xLab]

Problem

GeoSync already has numeric state quantization (analytics/signals/state_quantization.py) for continuous observations, but ORHSA/MVA release governance needs a separate claim-state quantum: an evidence compiler that maps incomplete, contradictory, local-only, CI-backed, or evidence-bearing proof chains into exactly one discrete claim state.

This prevents prose such as "almost ready", "mostly verified", or "green enough" from promoting a repository claim beyond its actual proof chain.

Existing foundation

• Existing module: analytics/signals/state_quantization.py
• Existing boundary: descriptor-only, not predictor / not financial advice
• Existing governance registry: governance/INSTRUMENTS.yaml

This issue must not duplicate numeric quantization. It defines the next layer: claim/evidence quantization.

Claim-state quantum

Allowed states only:

FALSE
UNTESTED
PARTIAL
LOCAL_VERIFIED
CI_VERIFIED
EVIDENCE_BEARING

Forbidden states:

almost ready
probably fine
green enough
looks correct
ship it

Quantization mapping

raw evidence -> normalized proof vector -> hard threshold gate -> discrete claim state -> allowed action

State semantics

• FALSE: contradicted by source, test, artifact, command, CI, or known negative evidence.
• UNTESTED: claim exists without executable proof.
• PARTIAL: some evidence exists, but one or more required links are missing.
• LOCAL_VERIFIED: source + tests + commands pass locally, with command output recorded.
• CI_VERIFIED: local proof plus same-SHA CI pass.
• EVIDENCE_BEARING: real data + replay + baseline + falsifier + semantic validation + provenance artifact.

Required evidence chain

Every claim must bind:

claim -> source -> test -> command -> artifact -> CI SHA -> failure mode -> rollback

If any link is missing, promotion is blocked.

Proposed artifact surface

• schemas/governance/claim_state_quantum.schema.json
• tools/governance/quantize_claim_state.py
• tests/governance/test_claim_state_quantum.py
• docs/governance/claim_state_quantum.md
• optional ledger integration: negative_evidence_ledger.md

Falsification targets

The implementation must fail closed when:

1. PARTIAL is promoted to READY.
2. local verification is claimed without command output.
3. CI verification is claimed without same-SHA CI.
4. evidence-bearing status is claimed without real data, replay, baseline, falsifier, and semantic validation.
5. a contradiction exists but the state is not FALSE.
6. rollback path is absent.
7. claim boundary is missing or implies physics/trading/predictive validity without validation artifact.

Acceptance gates

python tools/governance/quantize_claim_state.py --input <claim-evidence.json>
pytest -q tests/governance/test_claim_state_quantum.py
python -m compileall tools tests schemas

Expected gate behavior:

• incomplete chain -> PARTIAL
• no proof -> UNTESTED
• contradiction -> FALSE
• local proof only -> LOCAL_VERIFIED
• same-SHA CI proof -> CI_VERIFIED
• real-data scientific evidence chain -> EVIDENCE_BEARING

Claim boundary

Allowed claim: the quantum classifies evidence completeness into discrete governance states.

Blocked claims:

• It does not prove physics validity.
• It does not prove predictive power.
• It does not prove trading readiness.
• It does not replace scorecard gates.
• It does not make a release ready by itself.

Merge rule for future PR

One PR only. No runtime physics changes. No market/predictive language. Merge only after same-SHA CI is green and the PR body matches the final HEAD.