Expressive humanoid robots' awkwardness makes people wary, new brain study shows

People become more suspicious of a humanoid robot that makes errors, especially when the robot is an expressive conversation partner.

A new study published in Science Robotics involved 50 participants who held conversations and made joint decisions with the commercial humanoid robot Pepper, which is designed to be expressive and recognize emotions. The robot sometimes gave sound advice, sometimes made conversational mistakes, such as interrupting or pushing illogical suggestions. For some participants, the robot was animated with gestures, eye contact, and nods; for others, it remained motionless.

The researchers measured brain activity, levels of the hormone oxytocin, self-reported trust, and the robot's influence on decisions. They found that when people interacted with an expressive robot that violated interaction norms, their oxytocin levels increased. However, higher oxytocin levels during such errors were associated with lower trust and less frequent acceptance of the robot's advice. Oxytocin, often called the "love hormone," tracked with suspicion rather than affection in this context.

Errors damaged trust and diminished influence whether the robot was expressive or not. Expressiveness did alter how participants' brains processed errors. The study used functional near-infrared spectroscopy, a portable sensor worn on the forehead, to measure brain activity while participants moved and talked normally. The researchers focused on two brain regions: the dorsolateral prefrontal cortex, which monitors uncertainty and norm violations, and the medial prefrontal cortex, which supports "mentalizing" or inferring others' intentions.

When an animated robot erred, participants seemed caught off guard and worked harder to make sense of the awkward situation. Activity rose in both brain regions, and they worked together more closely. This coordinated activity predicted the rise in oxytocin levels, which in turn predicted falling trust and less influence on behavior. In contrast, this coordinated brain activity was absent in participants who interacted with expressionless robots.

The findings have important implications for robot design. A common assumption is that lifelike, expressive robots earn more trust, protecting their "reputation" even when mistakes occur. However, this study suggests that expressive cues shift the perception of errors from technical malfunction to social violation, engaging the same brain machinery used to judge people.

Future research will need to test whether these findings apply to women, mixed groups, other cultures, and different robot designs. The brain sensor only measured the front of the brain, leaving deeper regions unexamined. The researchers also want to explore whether robots can repair trust after mistakes by acknowledging errors, apologizing, or signaling good intent, similar to how people handle awkward interactions.

This study was conducted by researchers from Drexel University, the United States Air Force Academy, and George Mason University: Hasan Ayaz, Ewart J. de Visser, Frank Krueger, and Yigit Topoglu.

This article is republished from The Conversation under a Creative Commons license. Read the original article.