Part IV · Chapter 38Version 1.1 · 5 November 202629 min read
Bias Detection in Raw Data
Key message
Bias in raw data can be detected before training by combining historical context, representation tests against population benchmarks, statistical disparity tests and proxy screens, run automatically, intersectionally and with explicit attention to sample size and multiple testing.
Five chapters or files are free. This one would be your sixth.
Leave an email address and every chapter and every PDF of the compendium opens, in this browser, for a year. Nothing else is asked, and whether we may contact you is your choice.