Reading some relatively boring election-related stuff from Andrew Gelman, I came across an older post of his titled Against Parsimony. That stuck out to me, since parsimony had long been emphasized as an aspect of good theories. Fewer moving parts generally means less probability of error (this is related to the conjunction fallacy) and in practice it is necessary to avoid curve-fitting. Having a simple theory which predicts out-of-sample data is generally taken as an indicator as finding something about reality, which is why physicists looking for the most fundamental laws put so much emphasis on “beauty” or “elegance” of theories/equations. Gelman took his title from a paper by Albert O. Hirschmann. I still haven’t finished “Exit, Voice and Loyalty” though since starting it a while back I’ve bought and got midway through Mark Kleiman’s “When Brute Force Fails” and just today started Mancur Olson’s “The Rise and Decline of Nations”, in the introduction of which he spends a lot of words on the importance of parsimony in explanations of the titular problem.

Advertisements