Teaching (and marking) writing used to be one of the more unpredictable bits of my teaching life. Will we learn anything? Will we have a good time? Still, in a way, I’m glad I said goodbye to the classroom before the onset of AI mania.
I recently read John Warner’s useful analysis of what might happen to student writing in the wake of these text generators, and it got me thinking. What would I do if I were to assess my students writing again in 2025? How would I make sure it all works, isn’t a waste of time – and is accessible at the same time?
Would love to hear your ideas here, too – my socials are below. Here goes (in no particular order, as ever):
- Iterations by default. This would probably be the first thing I’d insist on. I would want my students to expect seeing their own writing several times over the course of their study – to edit it, change it, expand it, and so on. Would it stop all GPT use? No, but it might discourage the temptation to do it “just once – just for this essay”.
- Files with lots of version control. There are plenty of ways to enable this, I won’t get into the technicalities. The main thing I’d want from this is an awareness – shared by students and tutors – that all changes are recorded and seen. So when a five-paragraph chunk appears in the file all at once – alarm bells go off.
- Assessment milestones tied to process, not product. In other words: if I expect you to hand in an essay in October, another in January, and another in April, then the temptation to use Spicy Clippy is strong and justified. If I expect you to have made 50 edits by October, 150 by January, and 350 by April – then we’re still getting the essays, but what we’re judging is the work.
- Always, always, always make assessed writing very personal. There is the tendency to go for “objective truths” and to remain “academic” while teaching some writing. These are useful enough, and I’ll get to them. For the assessed pieces, though, I would make sure that students’ experiences are the main story. Combine this with iteration, and version control, and it’s much harder to get an LLM to hallucinate your personal accounts so consistently.
- Start on day one, and give lots of time. If this process starts on the first day – and if learners know there is plenty of time for constant work on these – then the whiplash effect of rushing to meet deadlines is dissipated. I can’t say it won’t be there – I study myself these days, and I know how powerful procrastination can get! But if, on day one, you tell students it’s not about three essays but about constant edits, then the emphasis shifts.
- Live-coding sessions for collaborative academic writing. What about the “serious”, objective stuff that teaches students academic writing skills? Simple: I’d do all of this collaboratively, and in class. If you make this part of their homework, or worse yet – part of their assessment, then students will instantly see the opportunity of outsourcing this to the robots, and pounce upon it. If you make this part of collaborative workshops, then at least you see the work being done, and are able to focus on the skills in a less soul-crushing setting.
- Mandate a secret journal, and assess only snapshots of it. The expectation to control, share, and examine all writing is, for me, the darkest aspect of teaching writing in the age of digital surveillance. There are some things I wrote – some I’m writing still – which are just for me. And in this way, writing is what makes me stronger, safer, in tune with what I want to attune to. If I taught writing again, I’d want my students to at least try this for themselves. There would be a workshop on how to set this up – with plenty of accessible alternatives discussed. And there would be a final component to the assessment, where I’d want to see a handful of journal fragments, and a piece about what it did to the student. Beyond that, I wouldn’t care.
- Do not expect “improvement” on every piece. This was always a bit jarring when I taught writing for exam prep classes: the things my students wrote would always need to get better and better. More structure! More ambitious linking words! Better grammar! That’s all well and good, but there are other directions to take your writing, besides “better”. What if we gave students one piece (say, out of four) on which “improvement” wouldn’t be mandatory? What if they still had to iterate, but could instead make it weirder – or fresher – or scarier – or whatever the fuck they wanted? Personally, I’d be more motivated to work on this piece myself – instead of getting a GPT to do it for me.
- Encourage “non-standard” language use. If your students are multilingual, tell them they can mix and match languages when they write for you. If they grew up with accents, or slang, or even a handful of words which only their family shared – give these things a chance to shine. Why would I do it? Simple: by personalising the way we use languages, we make our writing more creative. And that is how we keep ahead of the machine variety.
- Get over “collusion”, and encourage your students to borrow each other’s work. Finally, an idea which might be counterproductive at first. Would I really want to let students copy each other’s ideas? Yes. If the alternative to turning to GPT is, for them, to turn to another writing human for inspiration, then I’d 100% want them to do the latter. And even if this is a false dichotomy – then I’d still expect my students to see writing as something social, and not isolated: as a conversation between people, and not a get-out-of-jail note to nobody on the wall of your echo chamber.
How is any of this accessible? Simple. By giving students time and flexibility, you let them manage their pressure and effort. By showing that you’ll expect to mark work, not product – you level the playing field for those who would otherwise be disadvantaged, when everyone else turns to inaccessible GPTs for output. By encouraging collaboration, you would also need to ensure that everyone is included.
Would you disagree with any of the above? Would you add your own ideas? Let me know – all my socials are below.
Vic Kostrzewski (cost-chef-ski, he/him) is a Learning Designer, Translator and Project Manager based in South Wales. To discuss a new project, email anytime: vic@cost-chef.ski
(Photo credit: Photo by Deb Dowd on Unsplash)
