Triadic Human-Robot Conversation for Easier Disclosing: A Case Study Involving Individuals with Autism Spectrum Disorder

Friday, May 12, 2017: 10:00 AM-1:40 PM
Golden Gate Ballroom (Marriott Marquis Hotel)
J. Shimaya1, Y. Yoshikawa1, H. Kumazaki2, Y. Matsumoto3, M. Kikuchi2, H. Ishiguro4 and M. Miyao5, (1)Osaka University / JST ERATO Ishiguro Symbiotic Human-Robot Interaction Project, Toyonaka, JAPAN, (2)Research Center for Child Mental Development, Kanazawa University, Kanazawa, Japan, (3)AIST, Tsukuba, Japan, (4)Osaka University / JST ERATO Ishiguro Symbiotic Human-Robot Interaction Project, Toyonaka, Japan, (5)National Center for Chid Health Development,, Tokyo, JAPAN
Background: Individuals with autism spectrum disorder (ASD) often find it difficult to disclose their concerns to their caregivers (e.g., clinical psychotherapists, special education teachers, etc.). Previous research indicates that robots might be an easy conversation partner for individuals with ASD (Shimaya et al, 2016). An important next step is investigating how disclosing to robots can be bridged to disclosing to their caregivers. Triadic conversation including a robot, an individual with ASD, and his/her caregiver is expected to provide the bridge. However, the presence of the caregiver might decrease the advantage of easy conversation provided by robots.

Objectives: We examine whether individuals with ASD disclose their thoughts or concerns in the triadic conversation involving the robot and their teacher. We also examine whether the teacher can discuss the disclosed contents with them face to face after the triadic conversation.

Methods: Three individuals (two are male) with ASD participated in triadic conversations with a female teacher and a small desktop humanoid robot “CommU”. Their ages ranged from 23 to 25. One of them had a relatively low IQ (< 40) and the other two had a relatively high IQ (> 70). The participants were chosen because their teachers had difficulty in understanding their thoughts or concerns. The robot was tele-operated by the first author. Its utterance was produced by using a keyboard interface, and its gaze, which indicated who it was talking to, was controlled by a GUI. The teacher was forbidden from speaking except when replying to the robot. A participant-specific undisclosed and interesting topic list (UDI-list), consisting of topics that the teacher was interested in but unaware of, was prepared. The conversation was conducted mainly between the robot and the participant, and the robot asked the participant questions about the topics in the UDI-list, whereas the robot-teacher conversation on the same topic was sometimes inserted.

Results: The average conversation time for the participants was 20.7 min (SD = 4.7). Approximately 86% and 14% of the conversation time was spent on the robot-participant conversation and robot-teacher conversation, respectively. No participant stopped the conversation until the robot suggested that he/she finish it. All of them disclosed at least one of their thoughts or concerns to the robot in the presence of the teacher, to whom they had never disclosed on the topics. Examples of disclosed contents were his/her anxiety about job-hunting, and their hobbies, etc. The teacher conducted additional conversation with one participant and could discuss the disclosed contents with her face to face.

Conclusions: It is important that the individuals with ASD disclosed their thoughts or concerns even though the teacher was included in the conversation. The result implies triadic human-robot conversation might be effective in getting individuals with ASD to disclose easily before their caregiver, which involves their chances of disclosing indirectly to their caregiver. Increasing the number and the period of the sample, as well as developing an easier interface for the robot tele-operation to manage triadic conversation are important issues worth considering in the future.