EAGER: Studying Emotional Responses of Children with Autism in Interaction with Facially Expressive Social Robots

最近的研究表明，与不与机器人互动的同龄人相比，自闭症儿童在与机器人互动时表现出某些积极的社会行为。这些调查表明，与机器人的互动可能是自闭症儿童康复的一种有前途的方法。尽管在文献中报道了积极的迹象和观察结果，但使用社交机器人进行自闭症治疗的研究仍处于起步阶段，在研究机器人治疗自闭症的有效性和临床影响之前，还需要进行更多的基础性和探索性研究。本项目探讨了自闭症儿童与面部表情机器人互动时情绪和行为反应的几个研究问题。 代表性问题包括：（1）自闭症儿童是否能像典型发育（TD）儿童一样识别表情机器人的面部表情？(2)机器人是否应该使用身体姿势和动作与面部表情相结合，以更好地向自闭症儿童传达情感？(3)自闭症和TD儿童在与人类和机器人互动时的生理反应如何？为了解决这些问题，研究小组计划招募一组被诊断为高功能自闭症和TD儿童的儿童与面部表情机器人进行互动。这些会议是视频记录和分析使用国家的最先进的自动计算机视觉算法的面部表情识别。儿童的生理和情绪反应也是使用皮肤电活动（EDA）传感器获得的。EDA信号被用来创建一个计算情感模型来描述和比较两组儿童在与机器人交互时的唤醒（兴奋）水平。 这项研究的结果可用于设计针对人类精神和社会残疾的机器人辅助治疗，例如为患有自闭症，抑郁症或注意力障碍的儿童设计干预措施。此外，情感模型可以影响实验非交互式协议中的情绪条件的测量。该项目产生的数据是其他研究人员进一步分析的丰富来源。

Recent research suggests that children with autism exhibit certain positive social behaviors when interacting with robots when compared to their peers that do not interact with robots. These investigations suggest that interaction with robots may be a promising approach for rehabilitation of children with autism. Despite the positive signs and observations reported in the literature, research on using social robots for autism therapy is in its infancy and more fundamental and exploratory studies should be carried out before one can study the effectiveness and clinical impact of robots for autism therapy.This project explores several research questions on emotional and behavioral response of children with autism when interacting with a facially expressive robot. Representative questions include: (1) Do children with autism recognize facial expressions shown by an expressive robot similarly to typically developed (TD) children? (2) Should the robot use body gesture and movement in conjunction with facial expression to better convey emotion to children with autism? (3) How do physiological responses of Autistic and TD children vary in interaction with a human versus a robot? To address these questions, the research team plans to recruit a group of children diagnosed with High Functioning Autism and TD children to interact with a facially expressive robot. The sessions are video recorded and analyzed using state-of-the-art automatic computer vision algorithms for facial expression recognition. Children's physiological and emotional responses are also acquired using Electro Dermal Activity (EDA) sensors. The EDA signals are used to create a computational affective model to describe and compare the arousal (excitement) level of the two groups of children in interaction with the robot. Results from this study can be used in designing robot-assisted therapy for human mental and social disabilities such as designing interventions for children suffering from autism, depression, or attention disorder. In addition, the affective model can impact the measurement of emotional conditions in experimental non-interactive protocols. Data generated in this project are a rich source for further analysis by other researchers.