So they selected phone screeners with the same bias as the on-site screeners? Shouldn't the aim be to reduce correlation in order to increase the signal? Doing

So they selected phone screeners with the same bias as the on-site screeners? Shouldn't the aim be to reduce correlation in order to increase the signal? Doing the same interview 5 times would make hiring even cheaper, but that isn't really the most important metric you should optimize for.