But seriously, I think it is probably a good idea, if you have to throw out 90% of applications to start, to do it randomly if the alternative is a dubious stat

But seriously, I think it is probably a good idea, if you have to throw out 90% of applications to start, to do it randomly if the alternative is a dubious statistical or ML-derived model. It would make me feel better from the point of view of hiring or as a candidate, not to have the risk of an unforeseen undesirable bias.