Goodhart's Law is just the folk-decision-science version of overfitting, right? Would I lose anything by getting rid of the concept?
-
-
Replying to @regretmaximizer
@___jsf overfitting applies to induction specifically, not really straightforward to generalize.
1 reply 0 retweets 0 likes -
Replying to @ProofOfLogic
@___jsf a remark about overfitting which doesn't generalize: "you typically need an exponential amount of data to support a model's size"
1 reply 0 retweets 1 like -
Replying to @ProofOfLogic
@___jsf (what are "data points" for general Goodhart's Law?)
1 reply 0 retweets 0 likes -
Replying to @ProofOfLogic
@___jsf Also, Goodhart's Law deals with agents, unlike overfitting. Game theory, exploitable vs unexploitable policies can come into it.
1 reply 0 retweets 0 likes -
Replying to @ProofOfLogic
@ProofOfLogic in the analogy I'm imagining, data points are # of observations (however those get defined) that went into making a rule1 reply 0 retweets 1 like -
Replying to @regretmaximizer
@ProofOfLogic I don't see why exponential data size requirements shouldn't get to apply here too, though it would be nice to make an example2 replies 0 retweets 1 like -
Replying to @regretmaximizer
@ProofOfLogic also, so, nobody has tried to generalize overfitting to game theory then?1 reply 0 retweets 0 likes
@___jsf Maybe it's been done - I'm not aware of it.
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.