For example you must never declare 'magic numbers' in code. Or you must always obey S.O.L.I.D. or get 100% TDD. There will be be people who believe in these dogmatically to the point they won't employ anyone who says different (it becomes an interview question).
I am not arguing that these are wrong!
I am arguing that they are not evidence driven (they cannot be, software is to complex, it is not a narrow experiment on a lab mouse). So they must be culture/preference/worldview driven.
When there is no evidence driven approach to 99% of your decisions on software it becomes: an art. And that is fine.
That said it might be possible to show evidence that an approach is good for your code base, for your team, as that is a more limited scope, rather than "in general".
What isn't fine is the number of overly confident global assertions we hear from software people about how to build software.