When “anonymous” is not enough: methodological issues and the safety of human subjects in social media research

Drawing on data from a previous study on slur reclamation practices on Twitter/X, as well as scholarly discussion of context collapse and digital research ethics, I discuss the need to (re)evaluate how scholars engage with, publish, and present searchable language data online. Even when a subject’s social media persona is not linked –by name, location, and other identifying information – to their offline self, the distinction between the two is increasingly thin. Harm done to a person online –through harassment, dogpiling, suicide-baiting, other emotional abuse, and doxxing, among other tactics – is also harm done to their offline self. This is an especially salient risk for social media users from vulnerable or marginalized communities. I argue that stricter methodological and ethical standards should be established for research on social media language data, and present strategies myself and others have used (in various combinations) to tackle this issue: quotation with informed consent; discourse tallying; data aggregation; and focus on (in)famous public figures and organizations. I discuss the drawbacks and advantages of these methods, supplying examples from my work on metalinguistic attitudes towards slur reclamation on Twitter/X.