My basic point is that whatever such methods yield in the way of predicting subsequent job performance, retention, or whatever other outcome one is interested in, for the majority of both candidates and hiring managers, if there is no human-to-human interaction at any point in the process, you can pretty much scratch face validity.
There is a phenomenological aspect to both competitions and selection that I think we too often overlook. I would be among the first to acknowledge that if you have been unemployed for a while, whatever gets you a paycheck is fine by you. But, barring those sorts of circumstances, candidates generally wish to be selected for who they are and what other humans can clearly see they are capable of. If there is no human witness involved, then they have not been “picked”; they have only been processed.