Here is how models usually - and almost always - behave when they run into vague terms in your prompts:
-
They pick the most widespread interpretation of the term across different people. “Most widespread” does not mean that many people actually use it. The most widespread reading might be used by just a fifth of people, say. The vaguer the term, the smaller the core group that shares its most popular interpretation.
-
If the model picked an interpretation other than yours, the answer will be irrelevant to you. Bad. And the conclusions will be wrong on top of that - because you asked inside your own context, and got logic and conclusions from a different one, about something else entirely.
-
If you believe that you and the model are working from the same definition of a vague term, while in fact you are not, then you will be misled and convinced that the model confirmed or refuted the thing you asked about. It did nothing of the sort.
-
And from there it’s simple - your worldview gains one more error in your head.
The more precisely your terms are chosen, the harder they are to read in different ways, the sooner you and the person you’re talking to end up in a single shared context, understand each other correctly, and reach conclusions that are genuinely about the same thing rather than about different things.