Summary:A new study using SoftBank Robotics’ humanoid robot Pepper reveals how emotional expression, artificial intelligence, and human brain responses influence trust in social robots.

 

 

---

As humanoid robots move closer to everyday use in homes, healthcare, and workplaces, understanding human trust toward artificial intelligence systems is becoming increasingly important. A new study published in Science Robotics has revealed that expressive robots may create deeper social reactions — including stronger distrust when they make mistakes.

 

Researchers studied human interaction with Pepper humanoid robot, an expressive humanoid robot designed to recognize emotions, communicate through speech, gestures, eye contact, and facial movements. Fifty participants held conversations and made joint decisions with the robot while researchers compared different interaction styles.

 

With expressive humanoid robots entering homes, nursing homes and work places, how people react to them is increasingly important.

 

The study found that when Pepper provided incorrect advice, interrupted conversations, or made illogical suggestions, participants trusted the robot less. Interestingly, expressive behaviors did not protect the robot’s reputation. Instead, human brains appeared to interpret these failures as social mistakes rather than simple technical errors.

 

Using wearable functional near-infrared spectroscopy (fNIRS) technology, researchers monitored brain activity during natural conversations. They observed increased activity between brain regions responsible for detecting uncertainty and understanding another person’s intentions when participants interacted with expressive robots that failed.

 

 

The research also measured oxytocin levels, a hormone often associated with social bonding. Higher oxytocin responses during expressive robot failures were linked to lower trust and reduced willingness to follow the robot’s recommendations, suggesting that biological responses may reflect social suspicion rather than attachment.

 

The findings highlight a major challenge for future humanoid robots and embodied AI systems: making robots more human-like may improve interaction, but it also raises expectations. Future robot designs may need advanced error recovery, explanations, and apology behaviors to rebuild human trust after mistakes.

 

As AI robots become part of daily life, balancing emotional expression with reliability will be critical for successful human-robot collaboration.