Comment by beaconstudios

the problem isn't necessarily with _accurate_ data and math, but with reductive statistics that paint with broad brushes. Statistics inherently remove nuance, which is fine when the nuance is unimportant to what you're measuring, but not when it's actually important.

The example about cancer doctors in TFA is perfect. "more deaths = worse doctor" is a poor metric, because advanced cases have higher deaths in general, leading to a disincentive to try to help people with advanced forms of cancer. That's a terribly perverse incentive, and one that should be avoided.

Fundamentally, a lot of this stuff comes down to a lack of nuance in metrics, leading to some nasty effects down wind.