Infants Totally Not Fooled By Robots

Lately I've been worrying about potential robot influence on babies during the coming human-robot global conflict, but a journal alert popped up in my inbox today that has allayed my fears :

Infants understand the referential nature of human gaze but not robot gaze

Abstract: Infants can acquire much information by following the gaze direction of others. This type of social learning is underpinned by the ability to understand the relationship between gaze direction and a referent object (i.e., the referential nature of gaze). However, it is unknown whether human gaze is a privileged cue for information that infants use. Comparing human gaze with nonhuman (robot) gaze, we investigated whether infants’ understanding of the referential nature of looking is restricted to human gaze. In the current study, we developed a novel task that measured by eye-tracking infants’ anticipation of an object from observing an agent’s gaze shift. Results revealed that although 10- and 12-month-olds followed the gaze direction of both a human and a robot, only 12-month-olds predicted the appearance of objects from referential gaze information when the agent was the human. Such a prediction for objects reflects an understanding of referential gaze. Our study demonstrates that by 12 months of age, infants hold referential expectations specifically from the gaze shift of humans. These specific expectations from human gaze may enable infants to acquire various information that others convey in social learning and social interaction.

Infants develop the ability to follow the gaze of adults around 9-11 months of age, and it's not a coincidence that soon after that they start to speak their first words. Being able to tell what an adult is looking at is a key component of figuring out what the hell those strange sounds the adult is emitting might mean. This study suggests that by 12 months of age, humans are not only able to tell that adults are looking at something intentionally, but they reserve this inference for humans specifically.