This past spring, I taught a class about the teaching of writing to a group of aspirational high school teachers. Young, inspirational adults. We talked frequently about AI tools as well as writing with/out them. During our last meeting, one student asked me what I would recommend teachers do about students writing with AI. A sort of “final advice” request. Below is a revised (certainly more coherent) version of what I said.

AI writing is deskilling students. Have conversations about deskilling

We need to have long, difficult conversations that AI might be deskilling an entire generation of students. These conversations need to happen with students and parents. The parents need to be involved. Teacher-led initiatives always need a large coalition.

Let me explain the deskilling problem as briefly as I can. My entire life, older generations always had one major weakness in terms of job competition: a lack of comfort with new technologies. That’s simply not true anymore, mostly because the technologies are so easy-to-use. When I was a kid, I often had to fiddle in a terminal to get my videogames working. Or learn HTML to make a website. None of that is true anymore, hasn’t been for years. AI chatbots, such as ChatGPT, are radically easy to use. There isn’t some secret to using them, even though corporations (and unscrupulous online gurus) act like there is.

GenZ now faces intense job competition from older generations who are no longer technophobes. In fact, GenX remains a generation of top programmers (they know the programming language C and *cough cough* maybe Fortran). Python is easy to learn if you know C. Not the other way around.

Writing with AI is the same way. If you already know how to write, using ChatGPT is easy to use. If you don’t know how to write on your own, then people who do will always outperform you with respect to AI writing tools. Always. As a result, when a person doesn’t learn to write on their own, they’re deskilling themselves from the job market. They won’t be able to outperform those who have the habit of perseverance learned from writing. Consider, for example, summarizing. The point of summarizing, ideally, is not to quiz students but to make them glean and subsequently distill important ideas from a reading. Written summaries aren’t meant to be valuable themselves. Rather, summarizing is meant to indicate having thought through some ideas, putting them through your mind’s meat grinder. It’s an exercise designed to make you think. In fact, summarizing should never be graded. That defeats the entire activity of summarizing.

Move writing assignments in-person, sometimes with computers, sometime with blue books.

One major response to AI writing is to move the bulk of writing back into the classroom. Have students write in class where we can watch them write, including how they write, including the apps they use (with AI or without), offering them real-time feedback. Real-time feedback requires an intensity from the teacher because you need to be “on” the entire time. The bright side here: that real-time feedback shows students you’re engaged.

I think there is a general dislike of moving writing to the classroom that stems from our valorization of the author composing in isolation. It hearkens back to that weird “writer in the garret” image. Many people mistakenly believe good writing only happens through some sort of mysterious muse or when a writer is inspired.

I disagree. Treat writing like any other skill: practice it in challenging circumstances. Treat it like free throws in basketball. Try different techniques. Think about your form, i.e., posture. Think about your writing’s form, i.e., your genre, sentence technique, deliberate word choice, and many other dimensions. Have students set a time slot where all they do is write (say 15-20 minutes). There is an element here that teachers need to practice writing as well, if only to practice what we preach. I write 2 hours Mondays-Fridays. I treat it like work because it is my work. And I tell my students about my own frustrations with writing. I often recount days when I wrote a negative number of words (revision days).

I think the “blue book mindset” receives the bulk of the criticism here. This mindset usually isn’t writing as much as it’s taking a test via writing. There are other ways to do blue book writing that is more helpful. Blue book writing, as a journal, allows students to keep track of brainstorming. It’s useful for record-keeping around freewriting or quickwrites (never grade these, just check them to make sure students are doing the writing). Part of my own knee jerk reaction is that these kinds of in-class blue book activities are juvenile, reserved for grade school. Perhaps. But perhaps, grade schoolers are onto something, with their love of story telling built around low-stakes, ungraded writing. The entire point for them is to practice writing.

Teachers could start using the oral exam more extensively.

Oral exams have become a standard suggestion for circumventing AI use. This strategy has serious potential because there are strong relational elements between writing and speaking. I wrote an entire book about this, one I’ve made publicly available. I see this response as taking a forking direction. First, the oral exam could literally just be an exam where a teacher asks the student questions. We do this for PhD exams. The other direction involves having students write essays, hand them in, and then teachers ask them questions about their writing. If students can’t defend (or even recall) their own claims from the essay, this type of exam will reveal those problems.

There are a couple of drawbacks here. The obvious one: oral exams are time-consuming. That’s historically why we turned to writing papers. I don’t foresee this working for classes with more than 15-20 students. But if teachers are both supposed to read a paper and then meet with students, that’s an extraordinary amount of work.

The second drawback is one I haven’t seen discussed much: Teachers aren’t trained to ask students exam-type questions about writing in real-time. I thus advocate for training teachers in asking challenging, probing interview questions about writing. A new skill of 21^st century teachers is likely being a good interviewer. These interviews should take the high-stakes form of talking, like the job interview. Not like podcasting because these have audiences and are too meandering. A corollary here might end up reuniting writing and oral education components.

Show students we actually read the papers. And design assignments with that in mind

We need to show students we are reading their papers. True, this indicates we care. But it also means that we take our jobs seriously by offering legitimate suggestions for improvement. This is the one I’ve adopted the most. I make it clear to students that I read their papers by offering a substantive mid-term letter that I wrote without ChatGPT (copy editing mistakes included, artisanal writing, if you will). It takes a ton of time. I take these letters, print them out on letterhead, and sign my name by hand.

Use AI tools as modeling audiences, situations, and genres

AI tools have been trained on the entire internet and, possibly, millions of books. They simulate cultures, people, situations, and genres. It’s all incredibly biased, but so is reality. I think “Big Tech” misses the true power of their technological terrors (Darth Vader voice) by framing the chatbots with branding around time saving and “write for me” slogan. If they really believed in their products, they’d be trying much harder to design a thinking machine that responds to writing, not write for people. There is untapped potential here. For me, it’s the most promising, technically and socially.