The “top offender” model is really just a new version of Wichita’s old, court-challenged gang list, EFF’s Matthew Guariglia told @Kansasdotcom. “Mathematics provides a veneer of legitimacy.” https://www.kansas.com/news/local/article290434454.html
@eff @Kansasdotcom Is it "just math"? Yes.
Is math unbiased? Also yes!
Can I trust *any cop* to use the data produced by this algorithm in an unbiased way?
<oh_hell_no.mp3>
Honestly, the harder we make a cop's job, the safer we are from their mistakes, and from their biases and other bad decisions.
Real criminals will always mess up enough to get caught at some point. Simple probability will always work against you.
@arandmoor @eff @Kansasdotcom The choice of algorithm, inputs and their weights can be biased too, or at least suboptimal. Even if it wasn't, it still takes the possibly biased historical sentencing as an input.
It's good they included score decay.
