EthiCompass Lab
Every rate we report comes out of a defined statistical instrument. We write each one down in full — what it measures, how it aggregates, what it deliberately leaves out — and publish it here, so you can follow a figure back to the counts it came from, and so can whoever audits it later.
The papers below are definitional work in a running series, published in full and open to argument.
2 papers
Adapts the Safety instrument to a detector that returns a yes-or-no decision, and declares in the open what the simplification costs.
The definitional paper of the series: how an imperfect judge's continuous score becomes a rate you can read longitudinally.