Correlation between outcomes and race, gender, etc. is NOT bias per se. And yet that's what almost all claims of discrimination by machine-learned models are based on.
-
I’m trying to get at - assume blacks have higher rates of recidivism than whites by 2x. If model predicts 3x rate for blacks, isn’t that bias? Shouldn’t it predict 2x? I know it’s not that simple but I imagine there are other cases like this because of poor data/feature selection
