IDL Research Standard
Multi-AI Discovery Protocol v1.0
A repeatable research methodology designed to minimise hallucination, confirmation bias, false consensus and premature claims of discovery.
Core principle
AI generates possibilities. Evidence determines what survives.
Evidence maturity scale
| Level | Meaning |
|---|---|
| IDL-0 | Question |
| IDL-1 | Hypothesis |
| IDL-2 | Evidence |
| IDL-3 | Prediction |
| IDL-4 | Test |
| IDL-5 | Replication |
| IDL-6 | Discovery Candidate |
| IDL-7 | Validated Discovery |
Operational rules
- Independent AI researchers work in isolation before comparison.
- AI agreement is never counted as independent evidence.
- Sources must be directly verified before they support a claim.
- Alternative explanations and falsifiers must be actively developed.
- Predictions should be registered before confirmatory testing where practical.
- Red-team review is designed to break the preferred conclusion.
- Important results require reproduction and, where possible, independent replication.
- IDL-7 requires validation outside the originating IDL analysis.
This page is the concise public version. The full protocol can be published here later as a permanent, versioned research standard.