What I worry about is mismeasurement. They seem to have a clever classifier, but how are they training it? (My concerns here are relative to the mismeasurement

What I worry about is mismeasurement. They seem to have a clever classifier, but how are they training it? (My concerns here are relative to the mismeasurement of employee productivity that arises from having humans doing the evaluations.)