This symposium is an extension of a plenary forum on generative AI (hereafter GenAI) held at the EATAW Conference at Zurich University of Applied Sciences in Winterthur, Switzerland, in June 2023. Since the conference, AI – particularly the large language models (LLMs) shaping GenAI such as OpenAI’s ChatGPT – continue to develop rapidly with extensive integration and usage across disciplines and career sectors with educational and societal impacts. Given these developments, we recognize the central role that writing instruction has in fostering critical literacies and engaged usage and, at times, non-usage of GenAI. Just as we have adapted our teaching and learning to other technological developments, so too are we now at a time of transition and adaptation. Our initial discussion at EATAW was wide-ranging, intentionally so because (1) there is so much to explore in relation to GenAI, and (2) the EATAW membership is diverse, coming from a range of academic backgrounds. Thus in our original plenary and here in this symposium we have raised issues ranging from specific pedagogical approaches to questions of program and institutional administration, to broader public issues and conversations about the relationship of humans to machines. Here in this written symposium we each raise a different issue related to GenAI and writing with the aim to foster dialogue and discussion about GenAI in writing-related contexts.

Linguistic modeling of the writing process has gained in importance in recent years. Existing models, both from a linguistic perspective focusing on syntactic analyses as used in natural language processing and from writing research, are insufficient to actually linguistically explain what authors do when writing and revising. Writing is linear in time, but writers are free to move to any point in the text produced so far whenever they want, thus producing specific parts (e.g., sentences) in a non-linear fashion. However, the final product is a linear sequence of sentences. We therefore can interpret writing texts as a sentence-driven process. In this new framework, this article proposes a model of the production of sentences during writing. This sentence-centric model builds on existing considerations of transforming sequences, bursts and revisions, and takes into account aspects of linearity and non-linearity on the sentence level. We present a working implementation (available as open source software) and show which information can be gained by the resulting analyses in a small case study.