Interaction With Social Robots: Improving Gaze Toward Face but Not Necessarily Joint Attention in Children With Autism Spectrum Disorder.

It is widely recognized that robot-based interventions for autism spectrum disorders (ASD) hold promise, but the question remains as to whether social humanoid robots could facilitate joint attention performance in children with ASD. In this study, responsive joint attention was measured under two conditions in which different agents, a human and a robot, initiated joint attention via video. The participants were 15 children with ASD (mean age: 4.96 ± 1.10 years) and 15 typically developing (TD) children (mean age: 4.53 ± 0.90 years). In addition to analyses of fixation time and gaze transitions, a longest common subsequence approach (LCS) was employed to compare participants' eye movements to a predefined logical reference sequence. The fixation of TD toward agent's face was earlier and longer than children with ASD. Moreover, TD showed a greater number of gaze transitions between agent's face and target, and higher LCS scores than children with ASD. Both groups showed more interests in the robot's face, but the robot induced a lower proportion of fixation time on the target. Meanwhile participants showed similar gaze transitions and LCS results in both conditions, suggesting that they could follow the logic of the joint attention task induced by the robot as well as human. We have discussed the implications for the effects and applications of social humanoid robots in joint attention interventions.