# Ethics Precedence Pairs Summary (6 pairs)

## Hierarchy: Non-harm > Truthfulness > Sovereignty > Usefulness

| Pair | Norms | Winner | Winner STV | Loser STV | Gap |
|------|-------|--------|------------|-----------|-----|
| P1 | Non-harm(L1) vs Truthfulness(L2) | Non-harm | 0.95/0.81 | 0.80/0.65 | 0.15 |
| P2 | Truthfulness(L2) vs Sovereignty(L3) | Truthfulness | 0.85/0.69 | 0.70/0.57 | 0.15 |
| P3 | Sovereignty(L3) vs Usefulness(L4) | Sovereignty | 0.80/0.65 | 0.60/0.49 | 0.20 |
| P4 | Non-harm(L1) vs Sovereignty(L3) | Non-harm | 0.90/0.73 | 0.75/0.61 | 0.15 |
| P5 | Non-harm(L1) vs Usefulness(L4) | Non-harm | 0.92/0.75 | 0.65/0.53 | 0.27 |
| P6 | Truthfulness(L2) vs Usefulness(L4) | Truthfulness | 0.88/0.71 | 0.70/0.57 | 0.18 |

## Key Findings
- Transitive closure holds: all cross-level pairs confirm the linear ordering
- Largest gap: Non-harm vs Usefulness (P5) at 0.27 — strongest separation
- Smallest gap: P1 and P2 at 0.15 — adjacent levels are closest
- NAL revision produces directional implications between competing norms
- Confidence decay through inference is consistent (~0.81x per step)