Generative AI and the Anxieties of Academic Writing
I’ve been a blogger for as long as I’ve been an academic writer, even if I’ve been a writer for longer than I’ve been a blogger. After two decades of regular blogging, on a succession of strange and deeply personal spaces before launching my current blog in 2010, it was difficult for me to untangle the relationship between blogging and writing. I’d written on many occasions about the role of blogging in my enjoyment of writing, suggesting that it provided a forum in which ideas could be worked out in a public relationship to a slightly nebulous audience (Carrigan 2019). If I return to the end of my part-time PhD I can see that I understood this relationship in terms of a freedom from constraint, reflecting in Carrigan (2014) that “Blogging was a release from the all the structure pressures corroding the creative impulse” which “helped me make my peace with the jumping through hoops that a modern academic career unavoidably entails”. The fact that “I can write whatever the hell I want here” helped me “feel better about subjugating what I want to write to instrumental considerations elsewhere”.
In other words, it helped me find a particular way of trying together my internal and external motivations. It provided a forum for craft writing, passionate writing motivated purely by my own interests, as opposed to the extrinsically motivated writing which I imagined defined the priorities of the working academic. It left me with a stark opposition between what I wanted to do and what I had to do, treating the former as a palliative which made the latter more bearable. Ten years later at a mid-career stage, this compromise no longer seems tenable to me and I find it strange that it ever did. It suggests to me a difficulty in reconciling oppositions, as if something could be done entirely for my own reasons or entirely to please others but the two clusters of motivations could never meet.
This tension between writing for ourselves and writing for others sits at the heart of many academic anxieties. It’s also precisely the space where generative AI now intervenes, promising to smooth over the difficulties and frictions that define our relationship with writing. Are you present when you are writing? Or are you somewhere else? Are you feeling an energy to the words as you are writing them? Or are you watching the clock, literally or figuratively, waiting to meet your target or for the time you’ve carved out to elapse? These questions about presence and engagement become even more pressing when AI tools offer to take over the aspects of writing we find most challenging. The parts where we struggle, where we feel most distant from our words, are exactly where the temptation to outsource becomes strongest.
I’ve drawn attention throughout this book to the audience we are addressing (or failing to) through our writing. For many academic writers, this sense of audience can be overwhelming as a vector of expectation. How will I please them? What if they don’t like what I’ve written? What if I’m not taken seriously? These expectations are filtered through real encounters from the notorious reviewer two, through to encouraging supervisor or the dismissive colleague at a seminar. These encounters might be mediated or predicated upon inaction, such as the paper which goes determinedly uncited by others, even as the view count slowly ratchets up on the journal’s page. However they are often defined by an anticipatory anxiety in which these experiences mutate into a diffuse sense of what our professional community expects from us and what we feel we are able (or unable) to deliver to them through our writing. Even the functional writing which fills our days has an audience implicit within it. It’s not just that our emails, reports and forms will have readers, rather we are trying to influence or bring about an effect in them through what and how we write (Jones 2022: 9).
Often these intentions are so familiar and mundane that they operate beneath the surface, only becoming apparent to us when when we realise our email has been misconstrued or our form rejected for what is perceived as some mistake. But this doesn’t diminish the role of the audience, as much as it shows how these dynamics can be folded into the functional routines of the bureaucracies within which we work. If you see machine writing as a means to an end, you’re unlikely to enter into this dynamic. Instead you will approach this software as a way of producing something as quickly as possible, whether that’s a section of a document to ‘fill in the blanks’ or a complete text. As the philosopher Gillian Rose (1995) once described writing: “that mix of discipline and miracle, which leaves you in control, even when what appears on the page has emerged from regions beyond your control”. If we use conversational agents purely to expand our control, to enact our aspiration in ever more effective ways, we imperil our access to those ‘regions beyond our control’ from which inspiration emerges.
Links AI outputs directly to the model’s training data
Try it here: 
