When Algorithmic Fairness Fixes Fail: The Case for Keeping Humans in the Loop

amarashar's bookmarks 2020-11-30

Summary:

To Shah, the problem of algorithmic fairness is most concerning when it leads to unfair treatment in the clinic. A recent paper by Ziad Oberyemer received a lot of attention for exactly this reason, Shah says. There, a healthcare provider had used a cost predictive algorithm to decide which patients should be referred to a special high-risk care management program. The algorithm was one that used historic healthcare costs to predict future healthcare costs (and did so in an unbiased way). But when the healthcare provider used future healthcare cost projections as a proxy for healthcare need, the impact of that usage led to unfair treatment: Black patients had to be a lot sicker than white patients before they received the extra care.

Link:

https://hai.stanford.edu/blog/when-algorithmic-fairness-fixes-fail-case-keeping-humans-loop

From feeds:

Ethics/Gov of AI ยป amarashar's bookmarks

Tags:

Date tagged:

11/30/2020, 09:46

Date published:

11/30/2020, 04:46