We’d expect low trusters to be more on the lookout for deceptive behavior and yet they don’t perform as well as those who are more trusting.
The results were also a preliminary indication that the techniques and algorithms described in the paper could be used to successfully produce deceptive behavior in a robot.
