Children Interpret Human Gaze but Ignore Robot Gaze

The study coordinated by Antonella Marchetti of Universita Cattolica tested how young children interpret gaze from humans versus a humanoid robot. Per Phys.org, the experiment involved Italian children aged 3 to 5 who watched short videos in which either a person or a humanoid robot looked at one of two objects. ScienceBlog reports the sample size was 58 preschool children. The robot used in the stimuli is identified as Robovie in press coverage. Across the reports (Phys.org, NeuroscienceNews, ScienceBlog), children consistently inferred a target object as the human gazer's preferred item, but they did not make the same preference attribution when the gazer was the humanoid robot. The sources also report that gaze-human or robotic-did not change the childrens' own object choices. The paper appears in the International Journal of Child-Computer Interaction (DOI: 10.1016/j.ijcci.2026.100822), per Phys.org.

The findings document a boundary condition for basic social-cognitive inference: preschoolers deploy referential gaze to attribute preference reliably for human agents but not for mid-range humanoid robots. Industry-pattern observations note that humanlike appearance alone often fails to trigger full mental-state attribution; prior human-robot interaction (HRI) literature emphasizes contingency, reciprocal interaction, and contextual cues as stronger drivers of social attribution than appearance or isolated signals. For practitioners designing embodied AI for children, the result underscores that single-signal mimicry, such as reproducing eye movement, is unlikely to substitute for richer, multimodal interaction when the goal is to evoke humanlike social inferences.

The paper intersects developmental psychology and embodied-AI design, informing both basic research on theory-of-mind emergence and applied HRI for educational or caregiving robots. The result is relevant to teams building robots intended to engage preschoolers, early-education researchers testing social learning with machines, and ethicists tracking how children anthropomorphize technology. Because the study sample is modest (58 children) and stimuli were video-based rather than live interaction, its external validity to deployed robots in naturalistic settings remains an empirical question rather than a closed conclusion.

Observers should look for replication with live, contingent robot behavior and larger, diverse samples; experiments that vary reciprocity (turn-taking), verbal labeling, and multimodal cues; and HRI work that measures not only attribution but downstream learning or trust. Reporting to date does not include verbatim quotes from the paper's authors in the scraped sources; interested readers can consult the published article (DOI: 10.1016/j.ijcci.2026.100822) for the methodological details and authors' own framing.