The field of social robotics promises robots that can interact with humans in a variety of naturalistic ways, due to their embodiment, considered form, and social abilities. For providing a satisfaction survey, when compared to a web-based form, a social robot is theoretically capable of providing some of the benefits of a face-to-face interview without requiring a human. In this paper we set up our social robot, Opie, with a dialog-enabled chat-bot to run a satisfaction survey using off-the-shelf technologies. We collected audio and transcripts during the interaction, and attitudes towards the survey after the interaction. Twenty-one participants were recruited for the study, each played two games on a tablet and answered survey questions to the robot and through an electronic form. The results indicated that while participants were able to provide answers to questions, none of the components of the robot were robust to all the different situations that emerged during the satisfaction survey. From these results, we discuss how errors affected survey answers (when compared to the electronic form), and attitudes towards the robot. We conclude with recommendations for a set of non-trivial abilities that are needed before social robot surveyors are a reality.

中文翻译：

---

设计满意度调查机器人的挑战：使用社交机器人对人进行调查

社会机器人技术领域的机器人有望凭借其体现，考虑的形式和社交能力以多种自然主义方式与人类互动。为了提供满意度调查，与基于Web的表格相比，社交机器人理论上能够在无需人工的情况下提供面对面访谈的一些好处。在本文中，我们将社交机器人Opie设置为具有对话功能的聊天机器人，以使用现成的技术进行满意度调查。我们在互动过程中收集了音频和笔录，并在互动后收集了对调查的态度。招募了21位参与者进行这项研究，每位参与者都在平板电脑上玩了两场比赛，并通过电子表格向机器人回答了调查问题。结果表明，尽管参与者能够提供问题的答案，但机器人的任何组件都不能对满意度调查中出现的所有不同情况都具有鲁棒性。从这些结果中，我们讨论了误差如何影响调查答案（与电子表格相比）以及对机器人的态度。我们以社交机器人测量师成为现实之前需要的一系列非平凡能力的建议作为结束。