Paper presented at the workshop:
Digital and sensory expertise: Thinking deskilling and reskilling across automation. Organized by Brit Winthereik & Minna Ruckenstein

Abstract: Writing is a distinctive human endeavour of making inscriptions that allow for communication across time and space. Being literate and producing and deciphering these means of communication has been a source of power and prestige across different human societies for thousands of years. Writing in this way is a key technique for institutions of education, and the teaching of writing concerns fundamental questions of how institutions govern the ways in which populations think and act.
The recent advent of ChatGPT and other AI language models challenges a series of fundamental premises and norms about the question of writing in educational institutions, such as who controls the infrastructures of knowledge production, and what counts as expertise.
The first responses to the broad accessibility of technologies automating writing concerned how to control and assess student learning: how to organize exams and avoid plagiarism. It has also created discussions over norms and bias and the kinds of knowledge that these AI language models are informed by, and the business models they are part of. A less discussed topic is how the automation of writing also fundamentally challenges established knowledge about how students learn to write and how they gain a deeper understanding and generate ideas through writing. The point of departure for this paper is how in higher education, research on writing instruction over the last 40 years has shown that writing is not merely a means of communication, but that the very act of writing can give way to processes of thinking and realization. I show how writing with the ChatGPT can redistribute the process of generating text and thus not only recreate the intimate space of writing often highlighted in scholarship on writing, but also reconfigures the kinds of expertise involved in the writing practice. Rather than realizing ideas through text, putting words together, a key part of a writing process becomes prompting and critically selecting the best text. This means that expertise here - or literacy - becomes less a matter of the skill of writing sentences and more of having an overview of content and fields to critically select and prompt anew. Pointing to responses that indicate how teachers at all levels seem overwhelmed by the skill sets they need to develop to organize teaching and assessment using the AI language models, I argue that these skills not only concern digital expertise, but also an overview of didactics, and the ability to be explicit about genres and style and thus values that are often tacit aspects of academic writing and reasoning within particular disciplines.