I just read a great article over by Ben Goldacre, at the Guardian (British newspaper, SOURCE )
Ben's Column, called Bad Science, is about how people misuse science, usually the media, but also the British government. The US is no better.
This particular article complains about the lack of randomized trials for government policies. Instead of finding out if something works scientificially, we let the program administrators tell us whatever they want to, and usually don't find out how they failed (or worse, 'succeeded' at the same rate as no policy), unless there is an investigation.
So I hereby propose a simple rule - every pilot project include a randomized trial to see if it works. By 'pilot project', I would include any project that is not funded well enough to cover all possible participants. That is, if you are not giving enough money to house all the homeless, but only enough to house 1/10 (or even 1/2), that counts as a pilot project.
At heart, this would be putting the science back into politics. We could finally learn if a syringe exchange program really cut down on blood borne disease, of as some would claim, increased drug use. We could actually find out if the athletic teams (or music programs) helped students or took up valuable time for more academic pursuits.
The funding for such trials could come from the National Science Foundation, and should be relatively minor, as compared to the cost of the programs themselves.