As the writing across the curriculum and writing center coordinator on my campus, faculty ask me how to detect their students’ use of generative AI and how to prevent it. My response to both questions is that we can’t.
In fact, it’s becoming increasingly hard to not use generative AI. Back in 2023, according to a student survey conducted on my campus, some students were nervous to even create ChatGPT accounts for fear of being lured into cheating. It used to be that a student had to seek it out, create an account and feed it a prompt. Now that generative AI is integrated into programs we already use—Word (Copilot), Google Docs (Gemini) and Grammarly—it’s there beckoning us like the chocolate stashed in my cupboard does around 9 p.m. every night.
A recent GrammarlyGO advertisement emphasizes the seamless integration of generative AI. In the first 25 seconds of this GrammarlyGO ad, a woman’s confident voice tells us that GrammarlyGO is “easy to use” and that it’s “easy to write better and faster” with just “one download” and the “click of a button.” The ad also seeks to remove any concerns about generative AI’s nonhumanness and detectability: it’s “personalized to you”; “understands your style, voice and intent so your writing doesn’t sound like a robot”; and is “custom-made.” “You’re in control,” and “GrammarlyGO helps you be the best version of yourself.” The message: Using GrammarlyGO’s generative AI to write is not cheating, it’s self-improvement.
This ad calls to my mind the articles we see every January targeting those of us who want to develop healthy habits. The ones that urge us to sleep in our gym clothes if we want to start a morning workout routine. If we sleep in our clothes, we’ll reduce obstacles to going to the gym. Some of the most popular self-help advice focuses on the role of reducing friction to enable us to build habits that we want to build. Like the self-help gurus, GrammarlyGO—and all generative AI companies—are strategically seeking to reduce friction by reducing time (“faster), distance (it’s “where you write”) and effort (it’s “easy”!).
Where does this leave us? Do we stop assigning writing? Do we assign in-class writing tests? Do we start grading AI-produced assignments by providing AI-produced feedback?
Nope.
If we recognize the value of writing as a mode of thinking and believe that effective writing requires revision, we will continue to assign writing. While there is a temptation to shift to off-line, in-class timed writing tests, this removes the opportunity for practicing revision strategies and disproportionately harms students with learning disabilities, as well as English language learners.
Instead, like Grammarly, we can tap into what the self-help people champion and engage in what organizational behavior researchers Hayagreeva Rao and Robert I. Sutton call “friction fixing.” In The Friction Project (St. Martin’s Press, 2024), they explain how to “think and live like a friction fixer who makes the right things easier and the wrong things harder.” We can’t ban AI, but we can friction fix by making generative AI harder to use and by making it easier to engage in our writing assignments. This does not mean making our writing assignments easier! The good news is that this approach draws on practices already central to effective writing instruction.
After 25 years of working in writing centers at three institutions, I’ve witnessed what stalls students, and it is rarely a lack of motivation. The students who use the writing center are invested in their work, but many can’t start or get stuck. Here are two ways we can decrease friction for writing assignments:
- Break research projects into steps and include interim deadlines, conferences and feedback from you or peers. Note that the feedback doesn’t have to be on full drafts but can be on short pieces, such as paragraph-long project proposals (identify a problem, research question and what is gained if we answer this research question).
- Provide students with time to start on writing projects in class. Have you ever distributed a writing assignment, asked, “any questions?” and been met with crickets? If we give students time to start writing in class, we or peers can answer questions that arise, leaving students to feel more confident that they are going in the right direction and hopefully less likely to turn to AI.
There are so many ways we faculty (unintentionally) make our assignments uninviting: the barrage of words on a page, the lack of white space, our practice of leading with requirements (citation style, grammatical correctness), the use of SAT words or discipline-specific vocabulary for nonmajors: All this can signal to students that they don’t belong even before they’ve gotten started. Sometimes, our assignment prompts can even sound annoyed, as our frustration with past students is misdirected toward current students and manifests as a long list of don’ts. The vibe is that of an angry Post-it note left for a roommate or partner who left their dishes in the sink … again!
What if we were to reconceive our assignments as invitations to a party instead? When we design a party invitation, we have particular goals: We want people to show up, to leave their comfort zones and to be open to engaging with other people. Isn’t that what we want from our students when we assign a writing project?
If we designed writing assignments as invitations rather than assessments, we would make them visually appealing and use welcoming language. Instead of barraging students with all the requirements, we would foreground the enticing facets of the assignment. De-emphasize APA and MLA formatting and grammatical correctness and emphasize the purpose of the assignment. The Transparency in Learning and Teaching in Higher Education framework is useful for improving assignment layout.
Further, we can invite students to write for real-world audiences and wrestle with what John C. Bean calls “beautiful problems.” As Bean and Dan Melzer’s Engaging Ideas: The Professor’s Guide to Integrating Writing, Critical Thinking, and Active Learning in the Classroom (Wiley, 2021) emphasizes, problems are naturally motivating. From my 25 years of experience teaching writing, students are motivated to write when they:
- write about issues they care about;
- write in authentic genres and for real-world audiences;
- share their writing in and beyond the classroom;
- receive feedback on drafts from their professors and peers that builds on their strengths and provides specific tasks for how to improve their pieces; and
- understand the usefulness of a writing project in relation to their future goals.
Much of this is confirmed by a three-year study conducted at three institutions that asked seniors to describe a meaningful writing project. If assignments are inviting and meaningful, students are more likely to do the hard work of learning and writing. In short, we can decrease friction preventing engagement with our assignments by making them sound inviting, by using language and layouts that take our audience into consideration, and by designing assignments that are not just assessments but opportunities to explore or communicate.
How then do we create friction when it comes to using generative AI? As a writing instructor, I truly believe in the power of writing to figure out what I think and to push myself toward new insights. Of course, this is not a new idea. Toni Morrison explains, “Writing is really a way of thinking—not just feeling but thinking about things that are disparate, unresolved, mysterious, problematic or just sweet.” If we can get students to truly believe this by assigning regular low-stakes writing and reinforcing this practice, we can help students see the limits of outsourcing their thinking to generative AI.
As generative AI emerged, I realized that even though my writing courses are designed to promote writing to think, I don’t explicitly emphasize the value of writing as mode of discovery, so I have rewritten all my freewrite prompts so that I drive this point home: “This is low-stakes writing, so don’t worry about sentence structure or grammar. Feel free to write in your native language, use bullet points, or speech to text. The purpose of this freewriting is to give you an opportunity to pause and reflect, make new connections, uncover a new layer of the issue, or learn something you didn’t know about yourself.” And one of my favorite comments to give on a good piece of writing is “I enjoy seeing your mind at work on the page here.”
Additionally, we can create friction by getting to know our students and their writing. We can get to know their writing by collecting ungraded, in-class writing at the beginning of the semester. We can get to know our students by canceling class to hold short one-on-one or small group conferences. If we have strong relationships with students, they are less likely to cheat intentionally. We can build these bonds by sharing a video about ourselves, writing introductory letters, sharing our relevant experiences and failures, writing conversational feedback on student writing, and using alternative grading approaches that enable us to prioritize process above product.
There are no “AI-proof” assignments, but we can also create friction by assigning writing projects that don’t enable students to rely solely on generative AI, such as zines, class discussions about an article or book chapter, or presentations: Generative AI can design the slides and write the script, but it can’t present the material in class. Require students to include interactive components to their presentations so that they engage with their audiences. For example, a group of my first-year students gave a presentation on a selection from Jonathan Haidt’s The Anxious Generation, and they asked their peers to check their phones for their daily usage report and to respond to an anonymous survey.
Another group created a game, asking the class to guess which books from a display had been banned at one point or another. We can assign group projects and give students time to work on these projects in class; presumably, students will be less likely to misuse generative AI if they feel accountable in some way to their group. We can do a demonstration for students by putting our own prompts through generative AI and asking students to critique the outputs. This has the two-pronged benefit of demonstrating to students that we are savvy while helping them see the limitations of generative AI.
Showing students generative AI’s limitations and the harm it causes will also help create friction. Generative AI’s tendency to hallucinate makes it a poor tool for research; its confident tone paired with its inaccuracy has earned it the nickname “bullshit machine.” Worse still are the environmental costs, the exploitation of workers, thecopyright infringement, the privacy concerns, the explicit and implicit biases, the proliferation of mis/disinformation, and more. Students should be given the opportunity to research these issues for themselves so that they can make informed decisions about how they will use generative AI. Recently, I dedicated one hour of class time for students to work in groups researching these issues and then present what they found to the class. The students were especially galled by the privacy violations, the environmental impact and the use of writers’ and artists’ work without permission or compensation.
When we focus on catching students who use generative AI or banning it, we miss an opportunity to teach students to think critically, we signal to students that we don’t trust them and we diminish our own trustworthiness. If we do some friction fixing instead, we can support students as they work to become nimble communicators and critical users of new technologies.
