Peer review infrastructure

Reviews that disagree are the ones worth reading.

Lens keeps dissent visible instead of averaging it away, and publishes reviewer calibration so a low score means something.

11%

of reviews on Lens record a formal dissent, against 3% on the platforms it replaced.

Internal, 2,140 rounds

6 days

median turnaround, unchanged after calibration publishing was introduced.

Internal, 18 months

2.1×

more third-reviewer assignments, because splits are detected rather than resolved by rounding.

Internal, 2,140 rounds

Three things it does differently

Publishes reviewer bias
Every reviewer's mean offset from the panel is visible on their profile. Nobody is penalised for it; everybody reads scores in context.
Refuses to average a split
When two reviewers are more than two points apart, the round does not close. It assigns a third.
Keeps the dissent in the record
Minority opinions are published alongside the decision, attributed after the round closes.