The early development of young children’s imitation of social robots

Abstract

It is predicted that robotic teachers will be increasingly integrated into the home and classrooms of young children. Investigating whether children can learn from robots, and how well they do so, is critical to our understanding of how well this new technology works as a teaching agent and hence the impact it might have on the way children acquire novel information. The aim of this thesis was thus to explore the nature and extent of young children's social learning from robots. Past research has established that, from the second year of life, children increasingly copy the object-directed actions of others with high fidelity. Further, young children have been found capable of acquiring new information from a range of robotic teachers. However, what has yet to be established is whether or not children will imitate (i.e., copy the behaviour of a robot), and if so, (a) what is the developmental trajectory of this imitative behaviour, and (b) how does children's imitation of robots compare to their imitation of humans. Across three studies, this thesis explores whether children will imitate robots, how well they will imitate robots in comparison to humans, and how the imitation of robots changes with age. The first study explored children's imitation of live robots, revealing that 1- to 2-year-old children will imitate a robot, but produce fewer actions than when they see a human model, providing the first evidence of what I term the ‘robot deficit.' Importantly, the robot deficit was found to diminish: (1) with increases in age; (2) with a reduction in task complexity; and (3) with increases in the child's level of engagement with the robot. A second study replicated and extended these findings, demonstrating a robot deficit for imitation in children aged 1 to 4 years, and documenting that this deficit diminished with age and decreased when task complexity was reduced. A third study investigated overimitation – the tendency of children to copy everything shown to them by an adult model, even actions that are obviously causally ineffective – and if this behaviour in 4- to 6-year-old children is impacted by having a robot in place of a human demonstrator. This study found that children overimitate robots less than humans, again suggesting a robot deficit that extends beyond traditional imitation and into overimitation. This thesis provides the most comprehensive investigation to date of 1-to 6-year-old children's social learning from a robot, in comparison to a human, documenting evidence of a robot deficit. The work presented here lays the foundation for further investigation into the complexity of children's learning from robots in the first years of life and highlights that while imitation of robots is possible, age, task complexity and the quality of engagement with a robot are all significant predictors of successful learning.