Evaluation Registries — Document Frontier

Origins: research/evaluation/03-evidence-registry.md; 04-hypothesis-registry.md; 05-experiment-registry.md; 06-failure-taxonomy.md

Knowledge and limitations

The registries make claims falsifiable and preserve limitations. Evidence is strongest for repository observations and incident occurrence, weaker for transfer and causal design. The hypothesis registry has no HY-E007, creating an identifier-integrity question.

Five highest-value opportunities

1. RFR-003: Test HY-E003/HY-E005 with expert labels.
2. RFR-004: Test HY-E001/HY-E002/HY-E008 on a representative suite.
3. RFR-008: Test HY-E009/HY-E010 using control exercises.
4. RFR-006: Test HY-E011 with observed all-in costs.
5. RFR-009: Validate registry identifiers and cross-references.

Challenge: “Leading” hypotheses risk anchoring. Pre-registration and blinded adjudication should precede new observations.