Lessons in linguistics with ChatGPT: Metapragmatics, metacommunication, metadiscourse and metalanguage in human-AI interactions
Marta Dynel
Abstract
This paper discusses metalanguage, metadiscourse, metacommunication and metapragmatics testifying to users' conscious awareness enacted in human-AI interactions, based on a corpus of posts sent to Reddit's r/ChatGPT. The emphasis falls on users' foci of attention as they perform linguistic tests on ChatGPT and on how the “meta” practices manifest themselves interactionally on the selected subreddit, where human-AI interactions are showcased for entertainment purposes. The findings suggest that engaging with AI reflects and, potentially, enhances language users' metalinguistic, metadiscursive, metacommunicative and metapragmatic awareness. This awareness is mirrored in ChatGPT's output, indicative of its previous human-assisted training. Additionally, this investigation demonstrates that, when acknowledged as one subject of study, the four “meta” concepts are intricately intertwined as they may co-occur and overlap.