Perhaps my engineering bias is showing, but why not explicitly teach
students about making assumptions:
1. Which quantities are likely involved in a given problem?
2. What assumptions are usually good in the "real-life" situations most
of us encounter on a day-to-day basis?
3. Which additional assumptions are we making in order to simplify the
problem, even if they make the problem no longer representative of what
happens in most "real-life" situations?
4. After calculating the desired quantity with the desired assumptions,
how would changing certain of these assumptions affect the outcome?
Then, on tests, ask students to explicitly state which assumptions
they're making, and include questions that ask what would happen if
certain assumptions were changed.