> In the real world it means treating a study on 31 American undergrads as scripture

My opinion is that scripture shouldn't be treated as scripture to begin with.

> This policy by science simply doesn't seem very effective.

As a tool for making policy decisions, science gives you access to solutions that might work based on the evidence. Now, that evidence may be completely bogus, so there is no getting around the need for evaluation. But saying that data is a bad basis for goal-oriented action planning is a self-defeating position.

Policy by science is not a meaningful concept on its own, policy action requires a component we haven't talked about yet: a value system. Science can give you the data, but your goals determine what actions should follow. You're certainly correct when you say that politicians are getting some cover from improper data, but overall I find they seldomly hide their motivations.