Part of CNS 7.1 / GCTS: Grounded Chiral Tensor Synthesis

GCTS Worked Example

This synthetic example shows the behavior GCTS is meant to make explicit. It is not based on an active investigation.

Scenario

A safety incident is reported at a facility. A visitor says a staff member saw the incident and that policy should have required an incident report. The facility produces a visitor roster but does not produce an incident report, medical referral, or supervisor review. A staff statement says no report was required because the event was minor.

Question:

Did a documentation-triggering safety incident likely occur?

Candidate Claim

c_1: A documentation-triggering safety incident occurred at Facility A on Date T.

Evidence Atoms

IDSourceContentQualityNotes
e_1visitor statementVisitor reports seeing the incident and staff response0.70Direct but single-source
e_2facility rosterVisitor and staff were present at the relevant time0.85Produced official record
e_3staff statementStaff characterizes event as minor0.55Potential institutional incentive
e_4policy excerptVisible injury requires incident report and supervisor notice0.90Strong rule evidence

Record-Access States

IDExpected recordDutyAccess stateProduction stateConfidence
r_1incident reportpolicy_requiredunknownnot produced0.75
r_2medical referralconditional_on_visible_injuryunknownnot produced0.62
r_3supervisor reviewpolicy_required_if_reportinaccessibleno response0.58
r_4visitor rosterordinary_adminavailableproduced0.90

Candidate Worlds

WorldDescriptionKey assumptions
W_AIncident occurred and report was expected but not producedvisitor reliable, policy applies, record absent/non-produced
W_BMinor event occurred and report duty did not triggervisitor partly reliable, staff framing reliable, policy threshold unmet
W_CNo documentation-triggering event occurredvisitor mistaken, staff statement reliable, no report expected
W_DIncident occurred and report exists outside current access pathvisitor reliable, policy applies, record inaccessible

Example Scores

WorldPosterior
W_A0.46
W_B0.24
W_C0.12
W_D0.18

Claim posterior:

P(c_1 | E,A,I) = W_A + W_D = 0.64

Strict support:

P0(c_1 | E) = 0.00

Confidence:

Conf(c_1) = 0.52

Output Status

c_1: record_contingent / plausible-to-probable

The system does not promote c_1 to strict proof because the expected institutional records have not been produced. It ranks c_1 as likely under the top worlds while marking the claim record-contingent because production of the incident report, medical referral, or supervisor review could materially change the ranking.

Audit Output

{
  "claim_id": "c_1",
  "text": "A documentation-triggering safety incident occurred at Facility A on Date T.",
  "status": "record_contingent",
  "posterior": 0.64,
  "strict_support": 0.0,
  "confidence": 0.52,
  "supporting_evidence": ["e_1", "e_2", "e_4"],
  "refuting_or_qualifying_evidence": ["e_3"],
  "record_contingencies": ["r_1", "r_2", "r_3"],
  "top_worlds": ["W_A", "W_B", "W_D"],
  "next_records": [
    "incident_report",
    "medical_referral",
    "supervisor_review"
  ]
}

What The Example Demonstrates

  1. GCTS can rank likely truth without strict proof.
  2. Missing expected records affect status through access-state logic.
  3. A produced roster supports presence but does not resolve the incident claim.
  4. A staff statement can reduce confidence without eliminating higher-posterior worlds.
  5. The output identifies the records that would change the claim status.
Step 9 of 11 in CNS 7.1 / GCTS: Grounded Chiral Tensor Synthesis