The arguments the article lays out in favor of testing early autonomous vehicles as campus shuttles are compelling, and the reasons against not nearly so.

We expect a driver to provide certain services only because they happen to be there; like an elevator operator, toll booth operator, or person directing traffic, there's really no problem operating without them. You could easily put a "Call" button in an autonomous shuttle (as is seen in elevators) to cover most of the corner cases that need human intervention.

I think this test market is a great plan.