It is a failure of imagination and methodology to claim that it is necessary to experiment on millions of people without their consent in order to produce good data science.