In this paper, we investigate the effect of robot social behaviors in privacy-sensitive situations that are designed to reduce concerns of users regarding privacy invasion. As social robots ought to spend considerable time with humans, there has been research on user privacy protection. However, previous work mostly focuses on technical approaches not to disclose private information such as video/audio recordings and personally identifiable information. Although social robots could protect the privacy of users using those technologies, the users may still feel uncomfortable if they do not notice such the protection techniques being held. We design and test several behaviors of a social robot that can help users understand the internal process of the robot protecting user privacy so that users could have less concerns about their privacy. We choose three categories for the behavior design based on previous studies in psychology and human–robot interaction. They are (i) the gaze of the robot, (ii) the distance between the robot and a user, and (iii) clarity in expressing intent of the robot. In each category, we design and test three behaviors including a baseline (default) behavior. We conduct user studies with 56 participants in two scenarios to find effective behaviors. In the two scenarios, participants are asked to change clothes and write personal information in the presence of the robot in the vicinity, respectively. From the result, we find that users feel more comfortable in privacy-sensitive situations if they observe the robot perfor-ming the behaviors respecting user privacy. The most effective behavior of the robot is shown to be turning around from the user and showing robot’s back. As the robot combines more behaviors (such as telling the user that video recording is suspended, moving away from the user, and then turning around), the concern of users tends to decrease.

在本文中，我们调查了机器人社交行为在隐私敏感情况下的影响，旨在减少用户对隐私侵犯的担忧。由于社交机器人应该与人类共度相当长的时间，因此一直在研究用户隐私保护。然而，以前的工作主要集中在不公开视频/音频记录和个人身份信息等私人信息的技术方法上。虽然社交机器人可以使用这些技术保护用户的隐私，但如果用户没有注意到这些保护技术，他们仍然会感到不舒服。我们设计并测试了社交机器人的几种行为，可以帮助用户了解机器人保护用户隐私的内部流程，从而减少用户对隐私的担忧。我们根据之前在心理学和人机交互方面的研究，为行为设计选择了三个类别。它们是 (i) 机器人的凝视，(ii) 机器人与用户之间的距离，以及 (iii) 表达机器人意图的清晰度。在每个类别中，我们设计和测试三种行为，包括基线（默认）行为。我们在两个场景中对 56 名参与者进行用户研究，以找到有效的行为。在这两个场景中，参与者分别被要求在机器人在场的情况下换衣服和写个人信息。从结果中，我们发现如果用户观察机器人执行尊重用户隐私的行为，他们会在隐私敏感的情况下感到更自在。机器人最有效的行为被证明是从用户身上转身并露出机器人的背部。随着机器人结合更多的行为（比如告诉用户视频暂停、远离用户、然后转身），用户的关注度趋于降低。