This sounds like a solution to something, but I think it's a separate problem, and it's also potentially an introduction to another problem. In fields that aren't constrained to a single right answer, students frequently learn to optimize for being interesting and creative over coming up with the best supported answer they can. To quote Dave Barry

Never say anything about a book that anybody with any common sense would say. For example, suppose you are studying Moby-Dick. Anybody with any common sense would say Moby-Dick is a big white whale, since the characters in the book refer to it as a big white whale roughly eleven thousand times. So in your paper you say Moby-Dick is actually the Republic of Ireland. Your professor...will think you are enormously creative.

This sort of thing is good for stretching students' creative muscles, but bad for preparing students to grapple with tasks like "read these arguments for opposing conclusions and try and determine which is actually true, given that we can go out and check the answer objectively," or "Try and figure out whether this proposed design would work."

There is a problem with that: It assumes multiple valid responses and deals too much with what 'discover' or 'america' means. It wouldn't work for the 'Why is this steel plate hotter on the side away from the fire' question.

Hermenueutics-like games are risky since I think they teach contrarianness, thinking up unique-non obvious answers w/o regard to correctness. They teach the kind of reason that is rightly accused of being able to argue for atrocities.

I think that's a kind of terrible lesson.

It might work well to come up with a whole bunch of questions that are not to trick-questiony, but in which guessing the password is spectacularly wrong.

For the sciences, a better method might be to set up things where you have to make simple predictions.