AI ethics isn’t just a question of the future. We have an algorithm called COMPAS recommending whether people should go to jail or get bail in Florida right now.
Most modern AI ethics debates turn on making algorithms more “fair” but what is fairness? Even people can’t agree on what it means so how do we take a slippery concept, open to wildly different interpretations, and implement that in real life?
This article by MIT Tech journalists Jonathan Stray and Karen Hao, lead writer for The Algorithm newsletter, shows why “fair” is even harder than we imagined by taking a hard look at the COMPAS system. It let’s you play will different sliders to adjust the way the system works.
Can you do a better job than an judge and the programmers behind the software?
Move the sliders for yourself and it’s not long before you realize that no matter where you put the line, you’re always making trade-offs and compromises. Those lines will mark the battle grounds of the future as companies, governments and people try to decide just what fair means in the very messy real wold.