Empiricism guided by theory is the only way humans make meaningful progress. However, you need to treat claims very carefully.

I have a banner on the wall in my lab, "Nothing looks more like a discovery than a mistake." When you find something surprising, most of the time it's a technical error. You should always be massively skeptical of your own work ... and pretty skeptical of a novel finding of anyone else's, because you don't know if they showed the same skepticism. 95% of science reporting is sloppy researchers reporting a mistake, and credulous reporters treating it like a breakthrough. (1/4) You can't even necessarily trust large bodies of work, because with some frequency, they turn out to be fraud, in the worst cases, or groupthink, more commonly. When "95% of scientists agree" on something, maybe it's because it's true ... but maybe it's because they never thought about it that hard, and maybe it's because if you disagree, you can't get funded. Eating cholesterol causing high cholesterol always sounded stupid to me - and it turns out nobody'd ever actually checked. (2/4)