3 Comments
May 11Liked by Ben Recht

Can you say more about how you find “All models are wrong, but some are useful" to be generally misused? I feel like I usually see it used more-or-less in accordance with the ideas in this post.

Expand full comment
author

It gets thrown around a lot to justify any assumption, no matter how laughable or non-empircial. I get particularly peeved about it in applied stats, like say when we fit and interpret the coefficients of a linear model.

More broadly, I don't like that "useful" is normative and often just signifies "used."

Expand full comment

Reminds me of Vaihinger's as-if fictions

Expand full comment