关键词: 社交机器人, 孤独症谱系障碍, 人机互动, 社交技能

Abstract: Autistic Spectrum Disorders (ASD): one of the core symptoms in children with ASD is impairment in social interaction. Early intervention is critical for the development of sociability in children with ASD. Traditional rehabilitation interventions have limitations, such as their time-consuming nature, high cost and the shortage of professional rehabilitation therapists. Social robots are a tool designed to help people improve their social skills through social interaction. Social robots are widely used in intervention studies of social competence in children with ASD, with a focus on realizing the effective interaction between robots and users. Current findings of intervention studies using social robots for children with ASD show widespread problems of complex types and inconsistent standards of use, which makes it difficult to evaluate the feasibility and effectiveness of using social robots to help children with ASD.
The feasibility and effectiveness of the use of social robots for children with ASD can be improved from the following four aspects. First, it is necessary to increase the consistency of the classification and usage standards for social robots. Starting from the two dimensions of flexibility and degree of humanoid appearance, this paper divides social robots into non-humanoid robots and humanoid robots and introduces the application scenarios of different types of social robots. Second, we reviewed the findings of studies on social robot preferences in children with ASD. Children with ASD have a preference for the shape and technology of humanoid robots. When the appearance of a social robot is close to the image of a real person, it may cause the "uncanny valley" effect seen in normal people. To avoid this effect, designers often create robots that do not look like real people. However, individuals with ASD may not have an “uncanny valley” effect but an “uncanny cliff”. Such people do not show negative emotions when robots become more humanoid to a certain extent, but more humanoid robots may trigger social avoidance in individuals with ASD and negative emotions may be induced when the robots are indistinguishable from real people. This reflects the idea that individuals with ASD may have a less sensitive perception of humanoid robots than their healthy peers. Finally, the low sensitivity of individuals with ASD to humanoid robots may be due to their simplifying the complexity of human facial expressions and body movements and suppressing the intensity of influx of social information. Therefore, the negative emotions of ASD individuals may be alleviated to some extent. Moreover, gender, intelligence, and symptom performance in children with ASD may have an impact on the effect of interventions involving social robots. Therefore, researchers need to consider fully the above factors when designing and choosing social robots. The feasibility of using social robots as a tool for intervention in social disorders shown by children with ASD can be understood from psychological theories such as "new ontological category (NOC)" and social motivation theory. Social robots are not only a new entity but also a social reward for children with ASD. Therefore, social robots may have a transitional role in interventions facilitating the social skills of children with ASD.
Social robots have shown some effects in improving the social skills of children with ASD. The effect of such interventions is mainly reflected in joint attention (JA), self-initiation, verbal and non-verbal communication, empathy and the improvement of motor imitation. By analyzing existing studies, this paper summarizes the advantages of social robots in improving child engagement, improving attention, enhancing positive emotional arousal and making it easier for parents to understand and cooperate with their children. However, the studies also have the limitations of the experimental environment and methods, such as vulnerability to interference by external factors, robotics restrictions, high cost and difficulty in generalization of learning skills.
In conclusion, interventions involving social robots for children with ASD may have some feasibility and effectiveness. First, social robots could be considered as a "transitional object" in the intervention. However, when assessing the feasibility of the intervention one needs to consider the preference of children with ASD for social robots and its influencing factors, and to carry out intervention research on the basis of further standardizing the design and selection standards for social robots. Second, through in-depth analysis of existing studies, we not only expand understanding of the use of social robots in improving social ability in children with ASD, but also provide further research for reference. In future, research on the use of social robots in children with ASD may involve exploring the characteristics of the human-machine interface, revealing the psychological process of multimodal and brain-science technology, and paying attention to the development of artificial intelligence technology to build a closed-loop system for social robots.

Key words: social robots, autism spectrum disorder, human-robot interaction, social skills