Recently there has been a different type of crisis rhetoric around education. This “the sky is falling” event concerns “the OpenAI, ChatGPT interface [that] is now capable of producing convincing (though uninspired) college student quality writing to just about any prompt within seconds,” explains John Warner.
The freaking out has been a tad bit extreme: Daniel Herman announcing, for example, The End of High-School English.
Let me emphasize first that if you are concerned about AI-generated writing by students, please prefer Warner’s analyses and his two excellent books on writing (I use one with my first-year writing seminars): my review of The Writer’s Practice, my review of Why They Can’t Write, and my post about my FYW students’ response to The Writer’s Practice.
Now let’s focus on the hyperbole and the seemingly very real threat that AI-generated writing will erase writing assignments in K-16 education.
First, like Warner, I say: It’s the End of Writing as We Know It (And I Feel Fine).
I have been writing about writing and the teaching of writing for decades, and I have consistently challenged traditional approaches to writing instruction that is template and prompt driven. But I have also challenged the over-reliance on computer programs and technology to respond to, evaluate, and police student writing.
This new development around AI-generated writing is simply an extension of the Turnitin.com problem.
Once again, technology is not threatening student writing or the teaching of writing in K-16 education. Technology is exposing the essential problems with student writing and writing instruction in K-16 education.
As I outline in my chapter, De-grading Writing Instruction: Closing the “Considerable Gap,” for De-Testing and De-Grading Schools, the history of writing instruction in K-16 education is primarily one of misguided instruction, assignments, and outcomes. Yet, I also note that with the rise of the National Writing Project (NWP) and a move toward process writing in workshop contexts, there was a brief period of hope in the 1970s when momentum shifted in writing instruction toward what many writers and educators recognize as authentic composition.
And then A Nation at Risk and the tidal wave of accountability driven by standards and high-stakes testing washed away that hope.
The accountability movement ushered in the rise of rubrics and set the stage for computer grading of the so-called basic skills needed by all students—reading, math, and writing.
The consequences of this shift to accountability, which essentially ended the era of authentic writing instruction, resulted in teachers who believed they understood more than ever how to teach writing, but were in teaching/learning situations that did not allow for much writing to be assigned or for students to produce substantial amounts of original writing (see Applebee and Langer).
For example, when I was a beginning high school English teacher, my home state of South Carolina was an early adopter of exit exams, including a writing section.
My high school quickly pivoted to teaching to the test (see Bracey about WYTIWYG: “What you test is what you get”) by training struggling students to write 3-5-3 essays—a 3-sentence introduction, a 5-sentence body, and a 3-sentence conclusion.
This template provided the minimum amount of writing to be scored proficient but also limited the space in which students could demonstrate “errors” (we learned that patterns in writing, not single instances, triggered low scores).
The result was the highest passing rates in the state and a generation of students who wrote incredibly vapid and brief “essays.”
Now, if AI-generated writing can produce passages or even entire essays that meet the expectations of assignments in K-16 education, we shouldn’t be flailing our arms and racing around in Apocalyptic panic because that is a signal that the type of writing students are assigned and the writing they are taught to produce weren’t very good to begin with.
None the less, there appears to be a technology antidote available to those prone to seeking out technology—How to Detect OpenAI’s ChatGPT Output, Sung Kim.
Just as I see no need for Turnitin.com (and the research also refutes the values in the program; see the end of this post HERE), I believe the very real threat of AI-generated writing in K-16 education can be both a welcomed end to bad writing instruction, assignments, and essays by students as well as an opportunity to implement writing practices that greatly minimize students wanting or needing to cheat (similar to how we should be approaching traditional plagiarism).
Here, then, are my recommendations for addressing the Brave New World of AI-generated student writing:
- A key problem at the core of student writing and teaching writing in K-16 formal education is that the assigning and teaching of writing has disproportionately been the responsibility of ELA teachers (disproportionately experts in literacy and literature) who have little to no experience as writers and woefully inadequate preparation to teach writing. So a first-step to addressing writing in formal schooling is to better prepare teachers as writers and writing teachers (again, we have a ready-made process for that in the now underfunded NWP model).
- Next, a key way to encourage student engagement in writing and learning to write is to de-grade the writing process. See posts on de-grading and De-testing and de-grading schools: Authentic alternatives to accountability and standardization. Students who cheat are often driven by fear of failure or their inability to manage deadlines and workloads. Creating supportive and low-stakes environments for writing is foundational to both students learning to write and high-quality original student writing.
- A subset of the above point, writing must be a process and conducted in workshop class sessions. Process writing means students must produce artifacts demonstrating brainstorming and pre-writing, drafting, research, and revision after peer and instructor feedback. Plagiarism and AI-generated writing thrive in one-shot writing assignments driven by prompts; process and workshop writing by students support original thinking and writing as well as artifacts of the type of writing students can produce.
- Begin any course that includes writing assignments by having students produce in class a writing sample followed by having them submit a brief writing sample out of class. These samples can provide evidence for their writing styles and abilities.
- Include direct instruction and conversations in class about Turnitin.com and ChatGPT as well as why students engaging in authentic learning trumps trying to fulfill assignments or achieve specific grades.
- Finally, re-evaluate all writing assignments for authenticity and value in the course. If students can succeed with AI-generated writing in an assignment, that is likely a signal the assignment is the problem.
The fatalistic response to AI-generated student writing does not upset me because I have been making the same arguments above decades before this occurred. As I have often explained, writing and teaching writing are journeys, not destinations.
The threat of AI-generated student writing is not the end of that journey but an opportunity to take the fork in the road that we have been ignoring for decades.
AI Isn’t The Threat to High School English. Censorship Is: Book Censorship News, December 16, 2022
Thomas, P.L. (2019). Teaching writing as journey, not destination: Essays exploring what “teaching writing” means. Charlotte, NC: Information Age Publishing.