Method
Scientific AI Validation Checklist
Builds a validation protocol for scientific AI, simulations, imaging models or analytical workflows.
Field scoring contract
| Field | Type | Role | Direction | Scored |
|---|---|---|---|---|
| Hypothesis or task | text | context | context | no |
| Dataset provenance | select | evidence_signal | higher_is_better | yes |
| Baseline comparison | select | evidence_signal | higher_is_better | yes |
| Leakage checks | select | evidence_signal | higher_is_better | yes |
| Ablations | select | evidence_signal | higher_is_better | yes |
| Uncertainty reporting | select | evidence_signal | higher_is_better | yes |
| Reproducibility artifacts | select | evidence_signal | higher_is_better | yes |
| External validity | select | evidence_signal | higher_is_better | yes |
Limits
- Preliminary output
- Human review required
- Not certification
- Domain-specific checks added; not certification
- Context text is not averaged into numeric scores.
- Outputs require source/evidence review before decisions.