When a young child watches someone look at a toy, their eyes light up with understanding—they know exactly what that person wants. But show them a robot doing the same thing, and something crucial breaks down. A child of 3, 4, or 5 will read intention in a human glance but remain fundamentally unmoved by the identical gesture from a humanoid robot, according to new research from Università Cattolica Milan.
This finding matters because children increasingly encounter AI-powered robots designed to teach, comfort, and support them. If robots cannot communicate through something as simple and universal as eye contact, the implications ripple across education, autism support, and the entire landscape of how machines will learn to engage with young minds.
The study, published in the International Journal of Child-Computer Interaction and led by Antonella Marchetti, Director of the Department of Psychology at Università Cattolica and CERITOM (Research Center on Theory of Mind and Social Competences Across the Lifespan), tested Italian children aged 3 to 5 years old. The experiment was elegantly simple: children watched either a person or a humanoid robot shift their gaze from direct eye contact (lasting 4 seconds) to focus intently on an object (for 6 seconds). Researchers then asked whether the children could identify which object the gazer "preferred."
The results were stark and telling. When a human looked at an object, children readily attributed preference to that person—they assumed the human liked what they were looking at. The same gaze from a robot triggered no such inference. The children could see the robot looking, but they didn't translate that signal into any meaningful understanding of what the robot wanted or preferred.
What makes this especially significant is what it reveals about how children's minds work. Gaze is one of the first and most fundamental signals of human communication. Infants recognize it before language fully develops. Yet a robot wearing the same optical configuration—cameras positioned in the same way, pointed at the same objects—cannot crack through this seemingly simple barrier. The robot's gaze remains inert, a gesture without content.
Professor Marchetti explains that this does not doom robots to irrelevance in children's lives. Rather, it demands that designers think far more richly about interaction. "Simply imitating a single human signal, such as gaze, in a robotic artifact is not enough to make it truly communicative in a child's eyes," she notes. "Designing robots and intelligent technologies for children requires richer, more natural, and developmentally appropriate interactions: made up of words, gestures, reciprocity, context, and shared presence."
The research was conducted in collaboration with scholars from Tokyo and Osaka, and colleagues Davide Massaro, Cinzia Di Dio, and Federico Manzi at Università Cattolica Milan. One additional insight emerged: even human gaze alone did not shift children's personal preferences. Watching someone look at something didn't make children suddenly want it themselves. The children understood the other person's desire, but their own minds remained unmoved.
For applications in autism spectrum support—where gaze and shared attention are foundational skills—these findings offer crucial guidance. Humanoid robots are increasingly studied as rehabilitation tools, but they must be designed with richer interaction patterns to truly connect with children navigating difficulties with social communication. The data suggests that presence, context, and reciprocity matter as much as the signal itself.