This paper explores peoples attitudes about a service robot using customer data in conversation. In particular, how can robots understand privacy expectations in social grey-areas like cafes, which are both open to the public and used for private meetings? To answer this question, we introduce the Theater Method, which allows a participant to experience a “violation” of their privacy rather than have their actual privacy be violated. Using Python to generate 288 scripts that fully explored our research variables, we ran a large-scale online study (N=4608). To validate our results and ask more in-depth questions, we also ran an in-person follow-up (N=20). The experiments explored social & data-inspired variables such as data source, the positive or negative use of that data, and whom the robot verbally addressed, all of which significantly predicted participants’ social attitudes towards the robot’s politeness, consideration, appropriateness, and respect of privacy. Body language analysis and cafe-related conversation were the lowest risk, but even more extreme data channels are potentially okay when used for positive purposes.