Since the test subjects were instructed by researchers to follow the robots, their loyalty to the robots might just be their loyalty to the researchers, in disguise. In that case this is just the appeal to (human) authority bias that's already well-established. A more organic study would have people just happen to run into the robot, with no prior instructions on what to do. Then it would try to persuade them to do stuff.
I didn't read the paper though so I might have missed something.