from Hacker News

A more equitable AI model for breast cancer risk assessment

by finphil on 11/29/23, 9:49 PM with 1 comments

  • by Bostonian on 11/29/23, 9:56 PM

    The focus should be on making the model as good as it can be for all races, not "equity". If for some reason the model is 60% accurate on blacks and 70% accurate on whites, raising the accuracy for blacks would increase both overall welfare and equity, but reducing the accuracy for whites would improve equity without helping anyone.