| name | confidence-scoring |
| description | SOP for assigning calibrated confidence scores to causal claims based on evidence quality and quantity. |
| execution | sop |
| used-by | evidence-weighing, counterfactual-reasoning |
Confidence Scoring
Assign calibrated confidence scores to causal claims.
Tool
CC file edit
Protocol
- Count supporting evidence (weighted by evidence hierarchy)
- Count contradicting evidence (weighted similarly)
- Calculate net confidence: support_weight / (support_weight + contradict_weight)
- Adjust for mechanism plausibility (+0.1 if strong mechanism, -0.1 if no mechanism)
- Update claim page frontmatter confidence field
HARD-GATE
Confidence must be between 0.0 and 1.0. Must cite the evidence used in calculation.
Yield
Returns: { claim: string, confidence: number, evidence_count: number }