R0002/2026-03-13/C002/SRC01/E03

Research	R0002 — Research Standards for AI-Assisted Writing
Run	2026-03-13
Claim	C002
Source	SRC01
Evidence	SRC01-E03
Type	Factual

Five Downgrade Criteria Confirmed

URL: Not captured — experimental run

Extract

GRADE defines five criteria for downgrading the certainty of evidence:

1. Risk of bias (limitations in study design or execution)
2. Inconsistency (unexplained heterogeneity of results)
3. Indirectness (indirect population, intervention, control, or outcomes)
4. Imprecision (wide confidence intervals)
5. Publication bias (selective publication of studies)

These five criteria are confirmed by both the seminal GRADE paper and the CDC implementation handbook.

Relevance to Hypotheses

Hypothesis	Relationship	Strength
H1	Supports	Strong — confirms five downgrade criteria
H2	Supports	Strong — structural detail is correct
H3	Contradicts	Strong — five criteria is confirmed, not wrong

Context

The five downgrade criteria have remained stable since the framework's introduction. They are consistently listed across all GRADE implementations and training materials. No source disputes the count or the criteria names.

Notes

None.