Reducing Stress by Bonding with a Social Robot Towards Autonomous Long-Term Child-Robot Interaction Mike Ligthart Delft University of Technology The Netherlands m.e.u.ligthart@tudelft.nl Koen Hindriks Delft University of Technology The Netherlands k.v.hindriks@tudelft.nl Mark A. Neerincx Delft University of Technology & TNO Soesterberg The Netherlands mark.neerincx@tno.nl ABSTRACT Pediatric oncology patients could beneft from bonding with a social robot and talking about their day in the hospital. With our research we aim to contribute to the development of a robot that is able to facilitate a child-robot bond autonomously and long-term. We propose to use robot-disclosure and a shared interaction history to create a child-robot bond where the child feels comfortable and familiar enough to talk about their day with the robot. ACM Reference Format: Mike Ligthart, Koen Hindriks, and Mark A. Neerincx. 2018. Reducing Stress by Bonding with a Social Robot: Towards Autonomous Long-Term Child- Robot Interaction. In HRI ’18 Companion: 2018 ACM/IEEE International Con- ference on Human-Robot Interaction Companion, March 5–8, 2018, Chicago, IL, USA. ACM, New York, NY, USA, 2 pages. https://doi.org/10.1145/3173386. 3176904 1 MOTIVATION Cancer is one of the most profound health threats in modern society. Treatments are often extensive and life changing especially for children and their families. Providing social support to prevent late adverse efects, such as post-traumatic stress, is an important part of the treatment process [7]. Social robots could prove to be a useful tool for child-life special- ists to ofer social support, especially when human presence is not possible, e.g. during radiation treatment. The frst research into us- ing social robots as a companion to reduce stress during treatment shows promising results. Hospitalized children are for example more eager to emotionally connect with a robot than a virtual char- acter [6]. Using a Nao robot to assist a psychotherapist resulted in a signifcant lower amount of stress for pediatric oncology patients than after a therapy session with just the psychotherapist [1]. It is important to note that these studies, just like many other human-robot interaction (HRI) and child-robot interaction (CRI) studies, are based on a one-time interaction with a Wizard of Oz (WoZ) set-up. With the proposed research we aim to work towards a system that can overcome the challenges of long-term [3] and autonomous [4] CRI. One of our goals is for children to form a bond and chat about their day with the robot over multiple interaction. Permission to make digital or hard copies of part or all of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for proft or commercial advantage and that copies bear this notice and the full citation on the frst page. Copyrights for third-party components of this work must be honored. For all other uses, contact the owner/author(s). HRI ’18 Companion, March 5–8, 2018, Chicago, IL, USA © 2018 Copyright held by the owner/author(s). ACM ISBN 978-1-4503-5615-2/18/03. https://doi.org/10.1145/3173386.3176904 2 PROPOSED RESEARCH AND RELATED WORK Talking about the day and chit-chatting with the robot serves multi- ple purposes. It has the potential to contribute to the social support strategies currently in place [7] and it is the social glue between other activities with the robot [5]. In this research abstract we pri- marily focus on the technical and evaluation side. For an ethical and co-design perspective please consult [9]. The question is how to shape the conversational behavior of the robot. In Social Penetration Theory (SPT) the metaphor of a layered union is often used to describe a person. Forming a bond happens by penetrating the layers step by step by sharing bits of personal information with each other [2]. In order for children to feel comfortable enough to share their experiences with the robot a bond needs to be created. A bond that only can be formed if the robot self-discloses as well [5]. This process spans multiple interactions. Creating a sense of familiarity is essential to maintain a child-robot bond [8]. 2.1 Mutual Self-Disclosure When children are incited to self-disclose personal information they view the robot more as a friend and are more willing to interact with the robot again [8]. When the robot self-discloses personal information the children feel more inclined to do the same, the social infuence of the robot is stronger and the children appreciate the robot more [10]. We propose to create an elaborate narrative for the robot where it has a personality, a role, a ftting backstory and an extensive repertoire of anecdotes, dialogues and short personal stories. This narrative will be tailored, e.g. in terms of used language and themes, in advance to the developmental states of children. For example, a more witty robot for older children. Content from the repertoire will be selected dynamically based on the (emotional) state of the children and the interaction history (see next subsection). This is our implementation of robot-disclosure that should allow children to share about their day reciprocally. To manage the dialogue between the children and the robot we will introduce the robot as a chitchat robot in training: łit will make mistakes and it will not understand you all the time but it likes to learnž. Together with a tutorial on how to communicate with the robot, we call this ‘getting acquainted’, we intent to set the right expectations. We opt for a tablet supported dialogue, where questions are for- mulated following a recognizable structure. Children will practice to answer in a certain way to łhelp the robot understandž. This way we can constrain the conversation up to a level where autonomous Session: Pioneering Workshop HRI’18 Companion, March 5-8, 2018, Chicago, IL, USA 305