People make mistakes in communication all the time. Other humans can notice and interpret around those mistakes, but computers can't. Humans can infer things th

People make mistakes in communication all the time. Other humans can notice and interpret around those mistakes, but computers can't. Humans can infer things that are missing or implicit in the data. Computers can't. So if we build a computer system on top of data that we _know_ will be poor, incomplete, and sloppy, and if we force reliance on that system, then that's a major problem. It's silly to assert "the computer didn't do anything wrong". The computer has no agency or will. It doesn't matter if we run perfect software against a bad dataset, or if we run perfect data through bad software. The only thing that matters are the results. The computer and the people are all parts of the same system.