The vibe I get is that it isn't character they're really trying to judge. They're really trying to calculate whether a person is palatable to downstream players. I think there's likely some market risk associated with various personalities and they need to address it somehow even if there's no quantitative approach.

A wild experiment would be to test their judgment system by planting a few charming serial killers in their "character judgment" sample pool. My hypothesis is that the charming serial killers would pass at a higher rate.