Evidence Grading
Lightweight confidence method for distinguishing observations, validated claims, hypotheses, and data gaps.
Evidence Grading
Confidence Labels
- High: supported by multiple independent evidence streams or strong methodological repair.
- Medium: supported by project outputs but still dependent on coverage, metadata, or interpretation caveats.
- Low: plausible synthesis, prediction, or direction that needs explicit testing.
Required Caveat Discipline
Pages with numbers, rankings, or intervention implications must carry source projects and caveats. Claims from high-stakes domains need adversarial or independent review.
Agent Rule
If evidence is mostly extrapolated, write a hypothesis or direction page, not a claim page.
Source Projects
Collections
No direct collection links.