It’s been a crazy past couple days, so I wanted to flag a policy development that folks in the AI Ethics community might have missed. The Trump administration has made an eleventh hour attempt at dismantling disparate impact liability.(1/3) https://www.washingtonpost.com/education/civil-rights-act-disparate-impact-discrimination/2021/01/05/4f57001a-4fc1-11eb-bda4-615aaefd0555_story.html

This is relevant to AI ethics since disparate impact applies to discrimination cases without explicit evidence of intentionality. If there is going to be antidiscrimination protection against biased algorithms, it will mostly be through disparate impact doctrine. (2/3)

While this regulation would apply specifically to DOJ’s grant-making power for federal programs and might not withstand legal challenge, the fact that the administration is trying to take down key civil rights protections right before the transition is concerning. (3/3)