How likely are humans to trust a robot, especially if that robot has the capacity to lie? This is the question George Mason University psychology doctoral candidate Andres Rosero is exploring in his research on human-robot interaction.
“As robots become more integrated in human society, their roles will transition into those of a collaborative nature rather than as tools used by people,” said Rosero, who works in Applied Psychology and Autonomous Systems (ALPHAS) Lab in the Human Factors and Applied Cognition Program. “In response, these robots must be programmed with enough social awareness to navigate complex interpersonal interactions to build relationships and maintain a positive interaction with the humans they share their environment with.”
Almost 500 participants took part in Rosero's study “Human perceptions of social robot deception behaviors: an exploratory analysis,” which he conducted with George Mason graduate teaching assistant Harris Kelly and psychology professor Elizabeth Phillips. This study gauges how likely participants are to accept a lie told by a robot in various situational contexts and is one of the three manuscripts that comprise Rosero's dissertation.
Participants ranked specific scenarios involving robot deception and evaluated how deceptive the lie was, whether they approved of the robot’s actions, and if the robot’s behavior could be justified. The study placed these scenarios into three different environments: medical, domestic, and retail.
“The goal is to examine human perceptions of AI and robot collaborations across different complex social interactions,” explains Rosero, who received an MA in psychology from George Mason in 2022. “This study is one of two that examine moral norms in robots and how in breaking these moral norms, humans accept and justify the robot's behaviors.”
Each environment presented a unique quandary for the participant. For example, in the medical environment, the robot lied to a patient with Alzheimer’s, telling the patient that her deceased husband was coming home. In the other environments, the robot was evasive rather than uttering false statements.
In the domestic environment, the robot did not reveal to the individual that they were being secretly filmed. And in the retail environment, the robot did not disclose that it was capable of completing an assigned task and didn’t require human assistance. Overall, participants widely believed the robot’s lie in the first scenario, but once the robot’s capacity for deception and manipulation was revealed, they became less trusting. They also were more likely to say that the robot’s behavior in the medical environment was justified than in the other two scenarios.
“Any technology that is designed with the intention of being utilized by people is an intersection of that technology and the humanities,” said Rosero. “Our challenge as psychologists is to design experiments that properly explore this interaction to provide direction to the development of these technologies in the real world.”
Technology’s ability to potentially conceal its true capabilities remains a concern of Rosero, as a focus on human-robot dynamics continues to fuel his research. "I am planning on defending my dissertation in the summer, and I am hoping to continue doing human factors research,” says Rosero. “My goal is to conduct impactful research that can assist people in their interactions with cutting-edge technology.”
When offering advice to students in an age of artificial intelligence and its advancements, Rosero recommends understanding “the practical applications of your work.”
“As an applied psychologist, one of the most important questions I ask myself is how my research can be applied in the real world,” he said. “It is important to understand how your work helps to build upon the current knowledge of the field and how it could impact the development of the larger technology in society.”