Science is awesome. It is also incredibly difficult and expensive.

I would love it if we did proper science with education. I would love it if we did it with software development. In neither case has anybody put forward sufficient resources to do real science in these realms. Not science sufficient to replace human judgment.

Until we can, I don't think cargo cult science is the way to go. Half-assed studies on software development methods, counting simple metrics for developers, measuring teachers through simple standardized tests: all cargo cult science.

For now, I'd much rather rely explicitly on human judgment. It's not as good as actual science, but it's much better than pseudoscience.