R0002/2026-03-13/C002/SRC02/E01

Research	R0002 — Research Standards for AI-Assisted Writing
Run	2026-03-13
Claim	C002
Source	SRC02
Evidence	SRC02-E01
Type	Factual

Three Upgrade Criteria Confirmed

URL: Not captured — experimental run

Extract

GRADE defines three criteria for upgrading the certainty of evidence (applicable primarily to observational studies):

1. Large magnitude of effect (also called "strength of association")
2. Dose-response gradient
3. Opposing plausible residual confounding or bias (all plausible confounding would reduce the demonstrated effect)

Relevance to Hypotheses

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

Context

The upgrade criteria naming varies slightly across sources. "Strength of association" vs. "large magnitude of effect" are used interchangeably. The CDC version uses the more specific formulations listed above. The count of three is consistent across all sources.

Notes

Upgrade criteria apply primarily to observational studies, not RCTs. This is a nuance not captured in the original claim but does not affect the accuracy of the count.