#QOTD:
"Models are false in the same way that bridges are false - it's a category error to even label them true or false." @rlmcelreath
-
-
Replying to @davidmanheim
@davidmanheim@rlmcelreath Models are false in the same way that bridges are flat. They clearly aren't, but they have varying degrees of.1 reply 0 retweets 0 likes -
Replying to @LifeOfMyMoon
@LifeOfMyMoon@rlmcelreath point was about mind projection fallacy; http://lesswrong.com/lw/oi/mind_projection_fallacy/ … - models can be wrong *for doing X*, not wrong.2 replies 0 retweets 0 likes -
Replying to @davidmanheim
@davidmanheim@rlmcelreath The *for doing X* is essentially always implied as *for being true isomorphism of real world process*2 replies 0 retweets 1 like -
Replying to @LifeOfMyMoon
@davidmanheim@rlmcelreath Models useful because they have no purpose other than truth-tracking. They make no recommendations as to actions.1 reply 1 retweet 1 like -
Replying to @LifeOfMyMoon
@LifeOfMyMoon@rlmcelreath I do of work with models that include utilities; I can infer utilities from data, though, so it can be normative.1 reply 0 retweets 0 likes -
Replying to @davidmanheim
@LifeOfMyMoon@rlmcelreath (But now this question is about how we limit the scope of what we call a model.)1 reply 0 retweets 0 likes -
Replying to @davidmanheim
@davidmanheim@rlmcelreath I'd argue models, as the name implies, should be restricted to concept of representationpic.twitter.com/apz64izPAk
1 reply 0 retweets 1 like -
Replying to @LifeOfMyMoon
@LifeOfMyMoon@rlmcelreath ES Quade (1972); So represent for purpose of choice; good model depends on which choice.pic.twitter.com/ANndxKM63r
1 reply 0 retweets 0 likes -
Replying to @davidmanheim
@davidmanheim@rlmcelreath Complicated way of stating that a function from choice to outcome should be modeled well. Still descriptive.1 reply 0 retweets 0 likes
@davidmanheim @rlmcelreath Interested in a choice? Make it a variable x. Find f(inputs, x) -> output. f is your model, x can be optimized.
-
-
Replying to @LifeOfMyMoon
@LifeOfMyMoon@rlmcelreath That's unreasonable; you never have perfect models, you make tradeoffs between accuracy about different things.1 reply 0 retweets 0 likes -
Replying to @davidmanheim
@LifeOfMyMoon@rlmcelreath If you need f(inputs, x, y, z) to optimize x, or y, your best model is different. (Like the optimizer's curse).2 replies 1 retweet 0 likes - 14 more replies
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.