|
|
|
|
|
|
by kepler1
953 days ago
|
|
|
Two (?) some years ago a review story about Oregon's decriminalization policy was here on HN and I said that I didn't think it was a good idea, and was kind of shouted down by people saying, "this is the only rational choice, to do what works and stop penalizing people which costs a lot and doesn't change their behavior". Well, you can make stepwise rational choices all the way down into oblivion and watching your cities and states fall apart. It may be kind and more rational, but you've removed disincentives to use drugs and find yourselves in a society where you've attracted and created the conditions for this behavior to continue. No matter what your policy stance on the potential solutions, it is not good to be having a society where people are using serious drugs without penalty. Oh well, lesson learned, especially for people who thought they were liberal. Unfortunately at great expense and pain. |
|
|
This is a good point. We often see this pathological behavior in training neural networks (HN's latest interest du jour). It's easy to find local minima and think this is the best it's going to get, even if the global minima is in a completely different direction. Given the general danger of a stochastic optimization algorithm which would see us changing the law to increase harm, perhaps stepwise optimization algorithms are not the way one should run a government, yet this seems to be the only accepted political methodology these days. If someone suggests something completely new, even if they have ample evidence or a good causal model, they are often ignored.