Gwern talks about the reasons why most policies that have been evaluated don't actually improve sociological problems. (poverty, dependency, mental illness, crime) I find it surprising they don't hurt either. Gwern hypothesizes that it might be sample bias or maybe the forces underlying these problems are more powerful than the current capabilities of our institutions to handle. (eg. age? genetics?).

It's pretty unsatisfying. I hear that giving aid to the poor is treating symptoms not underlying causes, but we may not actually know the underlying causes to begin with. I want to say maybe spend more money on sociological research instead of services to the poor? Is that just yet another untested "clearly obvious" mass sociological policy?