CS doesn't really classify as a "hard science" either, to be fair.

All my physics professors, for example, were anal retentive about error bars--you had to put your error estimates in the notebook in ink before you took measurements. If something was "better than expected" you were about to have a painful journey figuring out why.