Friday, March 27, 2015

Don't be too impressed

Any predictive model which postulates tragedies will regularly appear to be validated.

A more interesting question is, how overjoyed is the modeler when that seeming validation arrives? And the follow-up: could there be a correlation between such schadenfreude and the fact that such models continue having opportunities to be validated?

1 comment:

  1. As Hannah Arendt put it, it is pointless to argue with a Nazi, because it is like arguing with a murderer if their designated victim is dead or alive. They have already developed the theory of why the victim is dead, so there is only one thing that remains to be done...