How we review

Evidence needs labels, limits and dates.

A short method for keeping claims inspectable without pretending uncertainty can be compressed into one score.

Evidence classes

We use six controlled classes. Established means broad or replicated evidence supports the core pattern. Supported means credible evidence supports it with meaningful limits. Mixed means important versions, measurements or boundary conditions disagree. Contested means serious unresolved challenges remain. Domain-specific means evidence should not be generalized beyond a narrow task or setting. Concept is used for reviewed terms or measurement ideas that are not standalone bias mechanisms.

What the class does not mean

It is not a numeric truth score, a diagnosis, or a claim that every study agrees. The descriptive evidence status and qualification remain more important than the label.

Source selection

We prefer original studies, systematic or serious narrative reviews, meta-analyses and stable scholarly identifiers. A source is linked to the claim it supports; the project’s interpretation remains separate.

Review dates

Each reviewed entry carries a date. Current, due and stale states are generated from the public freshness policy. A stale review is still visible, but it is not silently presented as newly checked.

Decision Audit

Audit eligibility is independent from evidence class. The tool does not diagnose which bias a person has.