The method

How peer review works

One engagement, ten moves — from a consented panel to honest re-measurement. Follow the map: solid arrows hand off data, the dashed branch is the red-flag safety valve, and the curved arrow loops each round back to the next. Click any stage for its detail and sources.

Confident about what we won't say

Too few raters in a group?

We withhold that stratum rather than expose it. A confident “not enough to protect anonymity” beats a leaky readout.

A safety disclosure is not feedback.

It exits the measurement pipeline entirely and goes to a named human — we surface it, we never seek it, and we never let a model decide.

Raters genuinely disagree?

When a group’s views don’t cohere, we say “raters disagree” instead of averaging them into a fake middle.

Rounds not comparable?

If the instrument shifted or invariance fails, we say the rounds can’t be compared instead of reporting a change that isn’t real.

See the whole thing on real people.

Run a pilot engagement with a handful of volunteers.

Run a pilot engagement →