If you have any problems related to the accessibility of any content (or if you want to request that a specific publication be accessible), please contact us at firstname.lastname@example.org.
Exploring Human Compliance Toward a Package Delivery Robot
AuthorWashburn, Andrew Logan
AltmetricsView Usage Statistics
Human-Robot Interaction (HRI) research on combat robots and autonomous carsdemonstrate faulty robots significantly decrease trust. However, HRI studies consistently show people overtrust domestic robots in households, emergency evacuation scenarios, and building security. This thesis presents how two theories, cognitive dissonance and selective attention, confound domestic HRI scenarios and uses the theory to design a novel HRI scenario with a package delivery robot in a public setting. Over 40 undergraduates were recruited within a university library to follow a package delivery robot to three stops, under the guise of “testing its navigation around people.” The second delivery was an open office which appeared private. Without labeling the packages, in 15 trials only 2 individuals entered the room at the second stop, whereas a pair of participants were much more likely to enter the room. Labeling the packages significantly increased the likelihood individuals would enter the office. The third stop was at the end of a long, isolated hallway blocked by a door marked “Emergency Exit Only. Alarm will Sound.” No one seriously thought about opening the door. Nonverbal robot prods such as waiting one minute or nudging the door were perceived as malfunctioning behavior. To demonstrate selective attention, a second route led to an emergency exit door in a public computer lab, with the intended destination an office several feet away. When the robot communicated with beeps only 45% of individuals noticed the emergency exit door. No one noticed the emergency exit door when the robot used speech commands, although its qualitative rating significantly improved. In conclusion, this thesis shows robots must make explicit requests to generate overtrust. Explicit interactions increase participant engagement with the robot, which increases selective attention towards their environment.