[Research] Stratification in negotiation: Do we need self-reference safety during Intent phase?

## Research Question

CLAIR uses stratification (L0, L1, L2...) with Löb discount to prevent confidence bootstrapping in Belief DAGs.

Does Intent (the negotiation phase) need similar protection? Or is stratification only relevant for committed Belief?

## The Concern

In negotiation, we might have:
```
Intent A: "I propose X", confidence: 0.9
Intent B: "I believe Intent A is well-founded", confidence: 0.85
Intent C: "Intent B's assessment of Intent A seems correct", confidence: 0.81
```

This looks like self-referential confidence bootstrapping. Without stratification, can negotiation participants inflate confidence through mutual endorsement?

## Hypotheses

1. **No stratification needed**: Negotiation is ephemeral; only committed Belief needs protection
2. **Implicit L0**: All negotiation is L0 by definition; only post-agreement Belief can be higher-level
3. **Protocol-level protection**: Repair/clarify mechanisms prevent bootstrapping by forcing grounding

## Research Direction

- Model a bad-faith negotiation where participants mutually inflate confidence
- Test if Intent's repair protocol naturally prevents this
- Determine if stratification should be enforced during negotiation or only at commitment

---
*Research thread: Self-reference in negotiation*