Apocryphal story about Schrodinger: he used his cat thought experiment as a litmus test to determine whether or not he should continue working on QM. When he got the result he didn't want, he switched over to biology, and wrote a book that inspired Watson and Crick.

I think we answer those sorts of questions through being careful thinkers and collecting evidence. If quantum mechanics is correct, what is made incorrect by that statement? Why do we believe the things that QM suggests are incorrect?

The double-split experiment, if correct, shatters our beliefs about microscopic existence formed by our experience with macroscopic objects, and has a number of testable predictions. We can construct other experiments based on those predictions, and when we do them it turns out that the microscopic world and the macroscopic world actually behave differently, but in a way that is consistent instead of contradictory. It's bizarre but it's possible.

We can check why my priors for insanity are higher than my priors for magic. I have solid evidence that I am more likely to believe supernatural claims because of irrationality than because of rationality.

Did Thomas Jefferson have solid evidence that meteors didn't exist? No- and it looks like he recognized that. He just had enough evidence to consider that the possibility discoverers of meteors were lying was stronger than the possibility they were telling the truth. The Platypus is another great example- many naturalists had enough evidence to believe it was a hoax rather than a real animal.

But the evidence against meteors and platypi is fairly small, and can be overcome relatively easily. What about the evidence against supernatural causation? It seems like the statement "you are insane if you believe in supernatural causality" might be true by definition.

What are the consequences if human porntuition is so strong that they predict where it'll pop up 53% of the time instead of 50% of the time? Either that statistical significance needs to be statistically significant (hey, when I run the results 100 times, I get a result at the p=.01 level! Fascinating!), that there's some systemic error in the setup, or that causality isn't unidirectional. I don't think I can express how much evidence we have against the last proposition.