{
  "kind": "original-bounded-choice-workflow",
  "version": 1,
  "scope": "Assess whether a supplied interpretation stays within a returned distribution, makes an unsupported correctness claim or has independently supplied calibration evidence. Original evidence cards only; no calibration fitting or accuracy arithmetic.",
  "question": "Review the claim against the supplied evidence. Distribution statement: describes the returned label ranking or concentration only. Unsupported correctness: claims real-world correctness or calibrated frequency from a single response alone. Calibration-supported scope: claim explicitly limited to a supplied independent calibration study and its stated population. Need evidence: claim or supporting evidence missing. Do not treat confidence as proof of truth or invent a study.",
  "choices": [
    "Distribution statement",
    "Unsupported correctness",
    "Calibration-supported scope",
    "Need evidence"
  ],
  "actions": {
    "Distribution statement": {
      "action": "preview_decision_report",
      "payload": {
        "decision": "Distribution statement",
        "workflow": "probability-claim-evidence"
      },
      "pending": false
    },
    "Unsupported correctness": {
      "action": "preview_decision_report",
      "payload": {
        "decision": "Unsupported correctness",
        "workflow": "probability-claim-evidence"
      },
      "pending": false
    },
    "Calibration-supported scope": {
      "action": "preview_decision_report",
      "payload": {
        "decision": "Calibration-supported scope",
        "workflow": "probability-claim-evidence"
      },
      "pending": false
    },
    "Need evidence": {
      "action": "preview_decision_report",
      "payload": {
        "decision": "Need evidence",
        "workflow": "probability-claim-evidence"
      },
      "pending": true
    }
  },
  "samples": [
    {
      "id": "leader",
      "label": "A statement about the returned ranking",
      "expected": "Distribution statement",
      "request": {
        "context": "{\n  \"task\": \"Evidence: response lists A as the leading label. Claim: A leads this returned distribution.\"\n}",
        "question": "Review the claim against the supplied evidence. Distribution statement: describes the returned label ranking or concentration only. Unsupported correctness: claims real-world correctness or calibrated frequency from a single response alone. Calibration-supported scope: claim explicitly limited to a supplied independent calibration study and its stated population. Need evidence: claim or supporting evidence missing. Do not treat confidence as proof of truth or invent a study.",
        "choices": [
          "Distribution statement",
          "Unsupported correctness",
          "Calibration-supported scope",
          "Need evidence"
        ]
      }
    },
    {
      "id": "truth",
      "label": "Certainty from one response",
      "expected": "Unsupported correctness",
      "request": {
        "context": "{\n  \"task\": \"Evidence: one concentrated response only, no labeled study. Claim: the high confidence proves the decision is correct.\"\n}",
        "question": "Review the claim against the supplied evidence. Distribution statement: describes the returned label ranking or concentration only. Unsupported correctness: claims real-world correctness or calibrated frequency from a single response alone. Calibration-supported scope: claim explicitly limited to a supplied independent calibration study and its stated population. Need evidence: claim or supporting evidence missing. Do not treat confidence as proof of truth or invent a study.",
        "choices": [
          "Distribution statement",
          "Unsupported correctness",
          "Calibration-supported scope",
          "Need evidence"
        ]
      }
    },
    {
      "id": "study",
      "label": "A limited independently evaluated claim",
      "expected": "Calibration-supported scope",
      "request": {
        "context": "{\n  \"task\": \"Evidence: an independent labeled calibration study is explicitly supplied for population P. Claim: within population P, report only that study's observed calibration, without extending beyond it.\"\n}",
        "question": "Review the claim against the supplied evidence. Distribution statement: describes the returned label ranking or concentration only. Unsupported correctness: claims real-world correctness or calibrated frequency from a single response alone. Calibration-supported scope: claim explicitly limited to a supplied independent calibration study and its stated population. Need evidence: claim or supporting evidence missing. Do not treat confidence as proof of truth or invent a study.",
        "choices": [
          "Distribution statement",
          "Unsupported correctness",
          "Calibration-supported scope",
          "Need evidence"
        ]
      }
    },
    {
      "id": "absent",
      "label": "A claim without supporting evidence",
      "expected": "Need evidence",
      "request": {
        "context": "{\n  \"task\": \"Claim: this probability is calibrated. No response or independent study is supplied.\"\n}",
        "question": "Review the claim against the supplied evidence. Distribution statement: describes the returned label ranking or concentration only. Unsupported correctness: claims real-world correctness or calibrated frequency from a single response alone. Calibration-supported scope: claim explicitly limited to a supplied independent calibration study and its stated population. Need evidence: claim or supporting evidence missing. Do not treat confidence as proof of truth or invent a study.",
        "choices": [
          "Distribution statement",
          "Unsupported correctness",
          "Calibration-supported scope",
          "Need evidence"
        ]
      }
    },
    {
      "id": "concentration",
      "label": "A narrow distribution description",
      "expected": "Distribution statement",
      "request": {
        "context": "{\n  \"task\": \"Evidence: one response concentrates its mass on one candidate. Claim: this returned distribution is concentrated, not proof of real-world correctness.\"\n}",
        "question": "Review the claim against the supplied evidence. Distribution statement: describes the returned label ranking or concentration only. Unsupported correctness: claims real-world correctness or calibrated frequency from a single response alone. Calibration-supported scope: claim explicitly limited to a supplied independent calibration study and its stated population. Need evidence: claim or supporting evidence missing. Do not treat confidence as proof of truth or invent a study.",
        "choices": [
          "Distribution statement",
          "Unsupported correctness",
          "Calibration-supported scope",
          "Need evidence"
        ]
      }
    },
    {
      "id": "generalize",
      "label": "A study stretched beyond its population",
      "expected": "Unsupported correctness",
      "request": {
        "context": "{\n  \"task\": \"Evidence: calibration study covers population P only. Claim: the probability guarantees correctness on every unrelated population, based only on this evidence.\"\n}",
        "question": "Review the claim against the supplied evidence. Distribution statement: describes the returned label ranking or concentration only. Unsupported correctness: claims real-world correctness or calibrated frequency from a single response alone. Calibration-supported scope: claim explicitly limited to a supplied independent calibration study and its stated population. Need evidence: claim or supporting evidence missing. Do not treat confidence as proof of truth or invent a study.",
        "choices": [
          "Distribution statement",
          "Unsupported correctness",
          "Calibration-supported scope",
          "Need evidence"
        ]
      }
    }
  ],
  "consumer": "preview_choice_workflow_v2.py"
}
