Frequently I see posts/articles which say that computer algorithms are racist. When I bother to read them, the logic is fairly simply--garbage in, garbage out. The algorithms are being developed using the conventional wisdom of whatever the subject is, and the conventional wisdom is racist.
I don't challenge any of that, but I'd insist the question is: so what?
Usually I take the message of the pieces to be--toss the algorithm out, it's biased, racist, undesirable.
But the true question is one of comparison: will using this algorithm instead of the existing process mean less racist results? My guess is usually the algorithm is likely to produce more consistent results, and usually less racist results. I see that happening because the algorithm would replace a more haphazard, variable process which evolved over time, and because the algorithm is being developed by people who are generally younger and more "woke" than those involved previously.
The second key question is: if we start to use the algorithm how likely is it that the algorithm can be improved?
No comments:
Post a Comment