Project resource
Scientific AI review checklist
A structured record for checking environmental claims, sources and domain reasoning.
The CSV includes columns for availability, source/date and your notes. The checklist does not require you to send information to BGF.
Review context
- Intended task, audience and decision consequences.
- Approved input material and information-handling boundary.
- Output, prompt context, system version and review date.
Claim and source checks
- Consequential claims separated from descriptive language.
- Source existence, exact supporting passage and applicability.
- Units, denominators, dates and operating conditions verified.
Scientific evaluation
- Mechanisms, mass streams and system boundaries checked.
- Uncertainty, unsupported inference and missing evidence identified.
- Rubric, representative cases and reviewer disagreements recorded.
Decision and traceability
- Correction, supporting rationale and remaining limitation for each material error.
- Responsible domain reviewer and escalation conditions.
- Reassessment trigger after changes to model, inputs or workflow.
This is an initial information checklist, not a laboratory instruction, safety assessment or detailed design specification. Select parameters and sampling arrangements with the responsible specialists.