# FM9 Refinement v2.2: Revision Probability Spectrum
## Prompted by Jon Grove, April 12 2026

### Previous versions
- v1: Verify against earliest record (too rigid)
- v2.1: Distinguish static facts vs dynamic state (too binary)
- v2.2: Revision probability spectrum (current)

### Core principle
No knowledge is truly static. All beliefs are revisable (fallibilism).
But revision PROBABILITY varies enormously across claim types.

### Spectrum
- NEAR-ZERO revision probability: historical events occurred, laws of physics
  -> Anchor to earliest record. Drift from earliest = likely confab.
- LOW revision probability: who said what in a specific conversation
  -> Anchor to earliest but allow recontextualization.
- MEDIUM: interpretations, understanding, evolving analysis
  -> Check both earliest AND most recent. Look for coherent evolution.
- HIGH revision probability: preferences, ranks, current state, evolving labels
  -> Demand auditable transition chain. Most recent with chain wins.

### FM9 procedure (revised)
1. Classify claim by estimated revision probability
2. Select verification method from spectrum above
3. Apply appropriate check
4. If uncertain about classification, run BOTH anchor and chain checks

### Philosophical grounding
Pragmatic fallibilism: certainty is graded, never absolute.
Anchoring is a pragmatic tool calibrated to probability, not an ontological claim.