The experiment
Researchers had 50 participants hold conversations and make joint decisions with Pepper, a commercial humanoid robot designed to be expressive and recognize emotions. Sometimes the robot gave sound advice. Sometimes it made conversational mistakes — interrupting people or pushing illogical suggestions.
For some participants the robot was animated, using gestures, eye contact and nods. For others it stayed motionless.
The team measured four variables: brain activity, levels of the hormone oxytocin, self-reported trust, and the robot's observed influence on participants' decisions. The findings were published in the journal Science Robotics.
What the numbers showed
Errors damaged trust and reduced the robot's influence on decisions regardless of whether the robot was expressive. But expressiveness changed how the brain processed the moment.
When an animated robot made a mistake, activity rose in two prefrontal regions — the dorsolateral prefrontal cortex, which monitors uncertainty and flags broken norms, and the medial prefrontal cortex, which supports 'mentalizing,' the work of inferring what another party intends. Those two regions also began working together more closely.
That coordinated brain activity predicted a rise in oxytocin levels. Counterintuitively, higher oxytocin during an expressive robot's errors correlated with less trust and less willingness to follow the robot's advice. The hormone, the researchers concluded, was tracking suspicion — not affection. The coordinated brain activity was absent in participants who interacted with expressionless robots.
The design assumption that fails
A common assumption in humanoid robot design has been that lifelike, socially expressive robots earn more trust, which protects a robot's 'reputation' even when it makes mistakes. The study's authors state directly that 'research is beginning to show that that assumption is faulty.'
Their explanation: expressive cues appear to shift how people categorize a mistake — out of 'technical malfunction' and into 'social violation.' A motionless robot's error reads as mechanical. The same error from an animated robot engages the cognitive machinery people use to judge other people.
The study used functional near-infrared spectroscopy, a portable sensor worn on the forehead that tracks oxygen levels in the brain during normal movement and conversation — a methodological advance over conventional MRI, which requires lying motionless.
The authors note key limitations: all 50 participants were young men, only one robot design was tested, and the brain sensor reached only the front of the brain. Next steps include testing women, mixed groups, other cultures and other robot designs.
CEO Times take
Robots are moving into homes, hospitals and workplaces at scale. The capital flowing into humanoid robotics is real, and so is the productivity promise. But this study is a useful corrective to the marketing reflex that 'more human' always means 'more trusted.'
For enterprise buyers deploying robots on factory floors or in elder-care facilities, the design tradeoff now has hard neurological data behind it: a machine that looks and acts like a person will be judged like one. When it fails — and machines always fail eventually — the reputational cost is steeper. Free enterprise rewards clear-eyed engineering decisions over feel-good anthropomorphism. The market, as always, will sort it out.
