pull down to refresh
10 sats \ 5 replies \ @SimpleStacker 11h \ on: A Critique of Black Box Economics econ
My problem is with any kind of absolutism in claims regarding methodology. Methodology should be chosen to suit the task. Whether you want accurate predictions, interpretability, or something else, depends entirely on the task
What happens when your chosen task fails when a new situation arises for the black box test? Do you just say, oh, that didn’t work well, we need a new black box model?
It looks like you agree with Friedman, empiricism is an adequate method.
reply
No, because any analyst worth their salt understands that black box predictive models aren't going to be as accurate or useful in scenarios not seen in the training data. Thus, when faced with entirely new situations you'll rely less on a black box predictive model and more on deductive reasoning.
reply
OK, but it seems there are not a lot of analysts that are worth their salt, now-a-days in the state apparatus, which is our overlord. If their models were working worth a sh*t, we wouldn’t be having the problems we are having now. Especially, in the economic control agencies like the Fed and all of the other regulators.
reply
I'd argue though that the incompetence of government workers has less to do with their methodology and more to do with the incentive structures inside government
reply
Hmmm…… I would argue that the incompetence is due to BOTH their methods and their incentives. It is a systematic failure of all they are involved in. They were chosen specifically because they conform best to the criteria of the state.
reply