Insider Brief
- A Drexel University-led study found that people formed stronger initial bonds with expressive humanoid robots but reacted more sharply when those robots made mistakes, making failures more damaging to trust.
- Researchers measured brain activity, oxytocin levels, surveys and behavior in 50 adult men interacting with Pepper, finding that mistakes by the expressive robot were processed more like social violations than mechanical errors.
- The study, published in Science Robotics and funded by the U.S. Department of Defense Air Force Office of Scientific Research, found that robot influence over participants’ decisions fell by more than half once errors began, underscoring the importance of reliability alongside social expressiveness.
A Drexel University-led study found that people formed stronger initial connections with an expressive humanoid robot but reacted more sharply when it made mistakes, suggesting that making robots more socially engaging can also make failures more damaging to trust.
According to Drexel, the research, published in Science Robotics, was conducted with researchers from the U.S. Air Force Academy, George Mason University and the University of Southern California’s Institute for Creative Technologies. The study was funded by a U.S. Department of Defense Air Force Office of Scientific Research grant.
Researchers found that participants engaged more deeply with a humanoid robot that made eye contact, gestured, nodded and responded while listening than with a motionless version that said the same things. But when the expressive robot began making errors and breaking social norms, participants were less willing to trust its recommendations.
The robot’s measurable influence over participants’ decisions fell by more than half after it began making mistakes, according to the study. Reliability ultimately mattered more than expressiveness for both versions of the robot.
“Trust is not one thing you can capture with a single measurement, so we measured several at once,” said the study’s senior authoer Hasan Ayaz, PhD, a professor in Howley College School of Biomedical Engineering and Science. “We recorded brain activity, hormone levels, what people told us on surveys and what they actually did, all in the same session, while human participants sat across from the robot and talked with it. Each modality tells a different part of the story, and only together do they show how trust builds and how it breaks down.”
The Method
The study involved 50 healthy adult men who each spent about two and a half hours interacting with Pepper, a humanoid robot, in a Drexel laboratory. Half met an expressive version while the others interacted with a stationary robot, according to the university.
Pepper, unbeknownst to the participants, was controlled by a human operator following a script, allowing researchers to keep its spoken responses consistent. Participants completed two interactions without deliberate errors before a third session in which the robot began interrupting, giving irrelevant comments and offering illogical explanations.
To measure trust in several ways at the same time, researchers used wearable functional near-infrared spectroscopy, or fNIRS, to monitor activity in the prefrontal cortex, collected saliva samples to measure the hormone oxytocin, surveyed participants and tracked whether they changed decisions after receiving advice from the robot.
When the expressive robot made mistakes, participants showed increased activity in brain regions associated with social reasoning. The researchers interpreted that response as evidence that people were processing the robot’s behavior more like a social violation than a machine malfunction. The same pattern was not observed with the motionless robot, researchers reported.
Oxytocin levels also increased as the robots made mistakes, even as reported trust declined. The researchers said the result suggests oxytocin may sometimes act as a signal of social vigilance rather than simply indicating bonding.
What it Means
“A charming robot that slips up pays a steeper price than a plain one,” noted the study’s corresponding author Frank Krueger, PhD, a professor in George Mason University’s School of Systems Biology. “Expressiveness is not free. It buys you engagement, and it buys you fragility at the same time, and that is a trade-off designers should be making deliberately rather than by accident.”
Researchers said the results suggest designers may need to weigh expressiveness against reliability carefully. A robot that behaves more like a social partner may attract greater engagement, but the Drexel-led study indicates that the same behavior can make mistakes more consequential when the machine fails to meet the expectations it creates.
“Reliability has to come first,” added corresponding author Ewart J. de Visser, PhD, of the Warfighter Effectiveness Research Center at the U.S. Air Force Academy. “An expressive robot that is unreliable is not a safer robot, it is a more disappointing one, because expressiveness raises a bar the robot then fails to clear. If a system is going to act social, it had better be able to back it up. Robot design shouldn’t look at likeability in isolation. Likeability matters, but engineering a social robot must take into account neurobiology and psychology to maximize performance.”
Disclosures
In addition to Topoglu, Krueger, de Visser and Ayaz, authors on the paper include Shawn Joshi and Nina Rothstein, former doctoral students in Ayaz lab at Drexel; Adrian A. Franke and Xingnan Li from the University of Hawaii Cancer Center; and Jonathan Gratch from the University of Southern California’s Institute for Creative Technologies. Topoglu and Krueger contributed equally to the work.
Funding for this research comes from the U.S. Department of Defense Air Force Office of Scientific Research grant. The views expressed are those of the authors and do not reflect the official guidance or position of the United States Government, Department of Defense, United States Air Force, or United States Space Force. As the technology developer, Ayaz holds a minor share in the startup firm fNIR Devices, LLC that manufactures optical brain imaging sensors used in the studies. The authors report no other conflicts of interest.
