causal-necessity-testing

$npx mdskill add yogsoth-ai/de-anthropocentric-research-engine/causal-necessity-testing

PNS evaluation: for each causal claim, determine whether the cause is necessary, sufficient, both, or neither.

SKILL.md
.github/skills/causal-necessity-testingView on GitHub ↗
---
name: causal-necessity-testing
description: "Tactic: Extract causal claims, evaluate probability of necessity (PN) and sufficiency (PS) for each, classify into necessity-sufficiency quadrants."
type: tactic
used-by: [counterfactual-probing]
strategies: [necessity-sufficiency, structural-counterfactual, thought-experiment]
---

# Causal Necessity Testing Tactic

PNS evaluation: for each causal claim, determine whether the cause is necessary, sufficient, both, or neither.

## Orchestration

1. **causal-claim-extraction** extracts all X→Y causal claims from the artifact
2. **necessity-evaluation** asks: if X had NOT occurred, would Y still hold? (PN)
3. **sufficiency-evaluation** asks: if X occurred in isolation, would Y follow? (PS)
4. Classify each claim into quadrant:
   - PN high + PS high → INUS condition (load-bearing)
   - PN high + PS low → necessary but not sufficient
   - PN low + PS high → sufficient but redundant
   - PN low + PS low → spurious or decorative
5. **load-bearing-identification** synthesizes quadrant assignments

## Scoring

- PN and PS scored 0.0–1.0 (probability estimates)
- Threshold for "high": >= 0.7
- Threshold for "low": < 0.3
- Middle range (0.3–0.7): uncertain, flag for deeper investigation

## Subagents Dispatched

- causal-claim-extraction (claim identification)
- necessity-evaluation (PN scoring per claim)
- sufficiency-evaluation (PS scoring per claim)
- load-bearing-identification (quadrant synthesis)

## Termination Conditions

- All extracted claims evaluated within budget
- Early termination if INUS condition found and budget is S
- All claims score PN < 0.3 (no necessary factors found — conclusion may be overdetermined)
More from yogsoth-ai/de-anthropocentric-research-engine