This question comes up every year on CD.
John V-Neun wrote a
great long blog post about the question a few years ago.
I think John sums it up pretty well. We can train students to drive the robot. We look for the things we can't train.
We also usually use similar criteria to determine the human player; they are usually a team member who we see with driver potential in the future. Frequently they become a driver a year or two later.
We will never issue a scored 'driver test' of any kind.