Robots are enforcing rules. How they treat people matters.

From monitoring public spaces to responding to requests for help, security robots are beginning to take on roles once held by people. As they gain authority, researchers are asking what it might take for the public to trust them.

In a new study, University of Michigan School of Information doctoral student Xin Ye and UMSI professor Lionel Robert find that fairness plays a significant role in how people trust the authority of security robots. 

Conducting an experiment with 100 participants, Ye and Robert found that the way a security robot interacted with people and the consistency with which it applied rules increased people’s trust in it. Fair interactions also increased participants’ willingness to accept the robot as an authority figure. 

“In human-robot interaction research, people often study robots as assistants, collaborators or partners,” Ye said. “I’m interested in what happens when robots take on roles of authority—when they exercise social power over humans.”

Their paper, “Trusting Security Robotic Authority: The Impact of Interactional and Distributive Fairness,” draws on research about fairness in policing to examine whether people bring similar expectations to robotic authority.

The researchers tested two kinds of fairness: interactional fairness and distributive fairness. Interactional fairness concerns how a robot treats people, including whether it gives them a chance to explain themselves and explains its own decisions. Distributive fairness concerns whether a robot applies rules consistently across groups. 

“Robots could hold authority in many settings, but security is particularly consequential,” Ye said. “They’re already being deployed. The NYPD deployed a Knightscope robot in New York City, though it was later taken out of service. Deployments like that leave us with important questions about how these systems should be designed and how people respond to them.”

The study used a parking-lot security scenario, and Ye said responses could differ in settings with higher stakes. Still, the findings offer a starting point for thinking about how security robots should be designed as their use expands. 

“If we’re going to design and deploy robots in those roles, we need to understand that fairness matters,” Ye said. 

Ye's paper was a runner-up for the Best Student Paper award from the Human-AI-Robot Teaming (HART) Technical Group within the Human Factors and Ergonomics Society (HFES). 

RELATED 

Learn more about Xin Ye and Lionel Robert by visiting their UMSI profiles. 

Check out UMSI’s PhD in Information program and apply today!