Well not always
As a statistician I only expect the answer 4. I could, within the limits of sampling, measurement & rounding errors, accept almost any number as the estimate of the result of adding two to two
One of the most painful periods of my professional life was when I had constantly to explain to customers that a table which we regularly produced could, quite legitimately, show that 5+5=5. Or 10. Or 15
But even in the world of counting with real, hard, integer numbers we all use daily a system in which, say, 10+4 = 2
And we also use a system in which 27+5 can give the answer 1 or 2 or 3 or 4
