“Mistakes”?

Was it ever designed to attempt to tell the difference? Aren’t most packing “robots” just machines a basic set of sequences with maybe a sensor or two for the most egregious problems, like “there’s something here, do thing” vs “nothing here, don’t begin sequence”