Could Human-Robot Interactions Facilitate Joint Attention of Children with Autism Spectrum Disorder (ASD)?

Oral Presentation
Friday, May 11, 2018: 2:40 PM
Grote Zaal (de Doelen ICC Rotterdam)
W. Cao1, W. Song2, X. Li3, S. Zheng4, G. Zhang5, Y. Wu3, S. He2, H. Zhu6 and J. Chen7, (1)South China Academy of Advanced Optoelectronics, South China Normal University, Guangzhou, China, (2)Centre for Optical and Electromagnetic Research, South China Academy of Advanced Optoelectronics, South China Normal University, Guangzhou, China, (3)School of Psychology, South China Normal University, Guangzhou, China, (4)School of Physics and Telecommunication Engineering, South China Normal University, Guangzhou, China, (5)Caihongqiao children rehabilitation and service center of Panyu district, Guangzhou, China, (6)Child Developmental & Behavioral Center, Third Affiliated Hospital of SUN YAT-SEN University, Guangzhou, China, (7)School of Electrical Engineering and Computer Science, KTH Royal Institute of Technology, Stockholm, Sweden

Impaired social communication is one of the core symptoms of the autism spectrum disorder (ASD). Some studies have shown that the children with ASD might have developmental deficits on joint attention, which was one of the basic forms of social interactions. Meanwhile, in recent years humanoid robots showed a potential to be a social partner of ASD and were introduced to involve in interventions. Nevertheless, few researches have focused on the difference of interacting with human agents and robots within ASD population. It remains a question whether human-robot interactions facilitate joint attention of children with ASD.


To address the aforementioned question, the study aimed to investigate the difference between the ASD and typically-developing (TD) children when viewing human-robot interactions compared to human-human interactions. By combining a commercial humanoid robot NAO with the Tobii X3-120 eye-tracker system, this study measured the eye movement patterns towards the interactions between two figures, including two types, namely human-robot vs. human-human.


6 ASD (mean age: 4.84±0.41) and 12 age-matched TD children (4.63±1.05) participated in the eye-tracking tests. Video stimuli depicted the interaction between one female and one agent (a male or a robot). Each video clip contained four stages: Greeting (two figures say “Hi” towards the screen); Looking at each other (two figures looked at each other); Leading by gazing (two figures turned their head towards the common target); Leading by pointing (two figures moved their fingers to point towards the target). First fixation time, total fixation duration and fixation time percentage were collected to measure the attention distribution among different areas of interests (AOIs): head, body, hand, target and background.


Data analysis showed that participants fixated faster in the human-human interaction condition in the stage with leading by gazing, but the total fixation time was shorter in the human-robot interaction condition. On the other hand, in the stage with leading by pointing, the participants made faster fixation in the human-robot interactions, but the total fixation time was shorter during human-human interactions. Meanwhile, compared to the human-robot interaction condition, in the human-human interaction condition the participants spent less time on the agent’s head, but spent more time on head and hand of the agent in the stage with leading by gazing, and on the target in the stage with leading by gazing. Furthermore, ASD spent more time on the male body, TD spent more time on human hand in leading by pointing stage of human-human interaction condition, but TD spent more on human body in leading by gazing stage, more on human head in leading by point stage of human-robot interaction condition.


This research showed that the human-robot interaction might have a different feature compared to the human-human interaction. Although the robot could attract the attention of ASD children in some conditions, the children with ASD could not well understand and follow the intentions of social robots, which implied that the social robot might have difficulty to facilitate the education or intervention of ASD children.