Aug 22, 2017 · 1 min read
Reminiscent of George Box’s description of statistical (and scientific) models! Statistics are always wrong, but that’s not the point. They produce results that can be usefully employed to do things. The catch is to be cognizant of the ways that models are wrong so that you don’t do silly things by taking them too literally. This, in turn, implies a paradox — you don’t “really” need experts in the real “science” behind the models, only those who have good enough understanding to make use of them (pilots, rather than aeronautical engineers), but you do occasionally need experts when you need to correct for overeager “pilots” taking things too far.