The way I think about it, it's all about levels of focus and asymmetrical loss of data.
If you zoom out to get more context you can't see all the details but vice versa (hence the motivation to have multiple levels of understanding) but the problem come when someone forgets there is a loss of data.
For example:
I don't know about cats, but I do know about mammals. I also know that cats are mammals. So if someone asks me about cats eg "do cats give birth to live young?" then I answer yes, because it is generally true of mammals even though I don't know if cats are an exception. And indeed this is going to serve me well in most cases. But if some asks the same things about monotremes, the "truth" I know about mammals is not true for monotremes.
And that isn't a big problem unless I decide that other people's right to live and marry and work and stuff is dependent upon them agreeing with my doctrine that monotremes are not really mammals. That's when the fights start!
With types of scale (compare interval to ratio) you can convert a ratio scale to an interval scale but not the other way around because ratio scales contain more data which an interval scale doesn't preserve.
Lables are a useful and necessary tool, but people are generally very unaware of their cognitive shortcuts and their shortcomings.