I think the obvious answer is, we can’t actually tell one way or the other if it is a net benefit. However, humans can’t handle uncertainty so must be told in confident tones that the decision is the correct one. If anyone questions the established decision, they must be discredited.
It's worse than that. We may know perfectly well which way is better on the net. But since masses can't handle uncertainty, you can't just say "doing this is more beneficial on the net" - you have to project certainty, or else people will interpret your lack of confidence as indicative of scheming, and do the opposite.
What related problem domains are understood “perfectly well”?
I think until we can accurately simulate the human body, there is essentially no hope for these areas. Unless it is possible to run highly controlled multi-generational studies with large numbers of people, statistical significance cannot be attained. Various meta analysis are funded and completed, but these do not count as strict definition science requires actual experiments to be independently replicated.
