In this article, we use grounded practical theory to develop a practical theory for using "difficult data" in Language and Social Interaction university classes. "Difficult data" are transcribed, audio/video-recorded data that contain language and ideologies that could be offensive, bigoted, or otherwise disturbing. We provide contextual literature for understanding the pedagogical tensions around using these data from communication pedagogy research as well as social theory for handling the language within the data. We base our analysis on various types of data from academia: an international/intercultural pedagogical speech event, student surveys, field notes, interview, and recordings of classroom instruction using difficult data. We provide a practical theory focused on techniques for managing the central problem that instructors want students to analyze the social consequences of interaction without replicating the problematic impacts of discourse that make difficult data difficult. Our theory reveals four techniques including framing the data as difficult, surveying students before the class, using already-published research data, and metacommunicating about these dilemmas. These techniques do not promise inviolability, but instructors can foster deliberative reflection/co-orientation about classroom interaction that considers whether the use of difficult data can help students develop analysis-informed praxis.