I think I first read someone bring up the “blue-book solution” for AI cheating shortly after ChatGPT exploded in fall 2022, but as I recall it, it was a joke. “Ha ha, now that AI can write as well as students, we’ll have to make them write by hand and while we’re watching. Ha ha!” My standard comment on social media to posts/articles about going back to handwritten and timed writing in the name of stopping cheating was “why not make them use a stone and chisel?” Ha ha.
Well, here we are three years in, and now blue-books really are a “solution” to AI. According to the Wall Street Journal, sales are up– way up. Earlier in August, Katie Day Good had an op-ed in The Chronicle of Higher Education titled “Bring Back the Blue-Book Exam,” and then at the end of August, Clay Shirky had an op-ed in The New York Times called “Students Hate Them. Universities Need Them. The Only Real Solution to the A.I. Cheating Crisis.” Both of these pieces make (mostly) serious arguments that the only way we can deal with/fight against AI cheating– a “crisis,” apparently– is to go back to the way we used to do these things. Way back.
Jeez.
Katie Day Good teaches at Calvin University in Grand Rapids and is “a media historian and cultural scholar of emerging technologies in education and everyday life.” A lot of her current work seems to be about “cultural movements to disconnect from digital technology and take a ‘digital sabbath,'” so maybe this return to handwriting is kind of in her research/scholarship lane.
But Shirky?!? Here’s a guy who became famous as a new media evangelist, who, in the book Here Comes Everybody, enthusiastically writes about crowd-sourcing everything and the joys of a world where content is both consumed and produced by users. His by-line describes his current job as “a vice provost at N.Y.U.” where he helps “faculty members and students adapt to digital tools.” This is the guy who is suggesting a return to blue-books and oral exams?!?
Jeez again.
Before I get more into the specifics of Good’s and Shirky’s essays, I want to bring up three “bigger picture” problems with blue-books and similar calls to return to the 19th century, problems that don’t come up in either one of these essays. First, blue-books, along with oral exams and other face to face assessments, obviously won’t work for an online class, especially ones that are asynchronous. And roughly speaking, a little over half of all college students take at least one class online, and about a quarter of all college students only take classes online. So what is an online teacher to do, collect blue-books by snail mail?
Second, timed writing like this is bad pedagogy, and people in writing studies have known this forever. No one is an especially good writer when they are being timed and watched, not to mention with no opportunity for things like feedback from peers or revision. I think these exercises are more like filling out a form than writing, and honestly, a better solution is some kind short answer/multiple-choice exam.
Third, and my apologies for offending anyone who thinks that blue-books might be a good idea, this is just fucking lazy. Good and Shirky are suggesting it’s just too much work for a teacher to change the assignment in some way where it is either not effective to use AI or that leans into AI in specific and useful ways. Shirky dismisses doing this work thusly: “We cannot simply redesign our assignments to prevent lazy A.I. use. (We’ve tried.)” It’s just too hard to do anything differently! Instead, Good and Shirky are saying we should travel back in time and just keep pretending that there is no other possible way to change how we do things.
I saw a version of this same logic at the beginning of my career in the early 1990s when word processing and internet technologies were emerging. There were similar efforts then to restrict student access to things like spelling and grammar checkers, or banning students from using online sources. Teachers– especially English teachers, I think– do not like to change how they teach, even when what and how they teach is altered by technology. As a result, teachers often follow the lazier solution, which is to ban the technology. Thus blue-books.
Both Good and Shirky begin the same way all of these AI freak-out essays begin: we can’t trust students at all and every one of them cheats on everything, especially now that it is so easy with AI. Good writes the new capabilities of AI made her rethink the “take-home essays” she used to assign in favor of blue book exams, presumably (in part) because of the possibility of cheating. Shirky begins with a vague story about a philosophy professor he met with who said he simply could not get “several” of his students to stop cheating with AI.
“Take-home essays” (I think she means what I’d call a take-home essay exam) have always required teachers to trust that their students won’t cheat. After all, when the student is working “at home,” there is nothing to stop that student from getting help from others and the internet, or even to get someone else to complete the assignment for them. I don’t know if Good was ever concerned about her students cheating on their take-homes before AI (she doesn’t seem to have been worried), but she started using blue books based merely on the possibility of cheating with AI.
As for Shirky’s philosophy professor colleague: I don’t know what several of them used AI to cheat means (are we talking half the class? three students? what?), but to me, the solution is obvious: fail them. I am going to assume (perhaps wrongly) that this hypothetical professor Shirky cites has a policy that does not allow students to use AI, and I’m also going to assume that the professor explained this policy and the consequences of using AI, which (again, just guessing) was failure. So, what exactly is the problem? If it’s that easy for the professor to catch students cheating, why not just enforce your policy and fail those students?
My own approach has been to be very up-front with students about what I think is and isn’t cheating with AI (and the short version is it is cheating if the writer directly copies/pastes AI output into something that the writer said they wrote). If I think a student is cheating with AI– which, for me, is based on my admittedly not perfect sense of what a particular student’s writing “sounds like,” and the document history of their Google Doc— I talk to them about it. In the last year and a half or so, I have had a lot more students cheating than I did before AI, meaning I’ve had to have a lot more of those uncomfortable conversations with cheating students. I give them another chance to do the assignment right and almost all of them managed to turn things around and pass the class just fine. I had a couple of repeat cheaters last year and I failed them on the spot.
In a post on Substack where she was explaining why she’s using AI detection software, Anna Mills described a confrontation she had with a student who adamantly denied he cheated with AI even though Mills is almost certain he did. After all, students also know AI is difficult to detect. I get it, and it can be hard to prove AI cheating. I’m sure I’ve had students who have managed to get away with some AI that I would have counted as cheating had I known. But every time I have had that “I think you cheated” conversation with a student, be it with AI or old-fashioned plagiarism, that student has confessed, often in tears.
As I’ve said many MANY times before:
- Most students do not want to cheat.
- Students cheat when they are failing and they are desperate.
- Students who cheat are not criminal masterminds and are easily caught.
- All that said, it does depend on what exactly counts as cheating, and I don’t think it is cheating if students use AI as part of their process.
Good views this return to handwritten essays as a “balm for my tech-weary soul.” She goes on:
My students’ handwritten essays brim with their humanity. Each page conveys personality, craft, voice, and a “realness” that feels increasingly scarce in our screen-saturated, algorithmically-distorted information environment. As such, handwriting accomplishes something greater now than ever before in education: It restores a sense of trust to the student-teacher relationship that has been shaken by AI.
In the next paragraph, she also brings up some of the other beliefs in handwriting’s “authenticity,” that handwriting helps people make better connections in the brain than typing, that it results in better notes, etc. Well, right before Covid struck, I was researching laptop/cellphone bans in f2f classes and requiring students to take notes by hand. Long story short, the studies I’ve seen about comparing laptop notes with handwritten notes in classrooms– mostly quantitative/experimental methodologies coming out of Education/Psychology– strike me as flawed for all kinds of different reasons. And the claims about handwriting as a tool for judging one’s “authenticity” and identity and the like have been debunked by many researchers– I would recommend in particular the very readable and well-researched book by Tamara Thorton, Handwriting in America: A Cultural History. I also have my own baggage as someone with terrible handwriting, who remembers failing handwriting in the fourth grade, and also as someone who has typed everything I could type since I was in high school.
So for me, the idea that handwriting is “better” and that it is both possible and reasonable to make judgements about the writer based on their handwriting, that more of one’s humanity is revealed through handwriting– that’s all bullshit.
Shirky doesn’t seem to think that handwriting has the same kind of “Magic” that Good sees in her students’ writing, and he admits that a lot of students and faculty are skeptical of this change. But in the name of rigor and a “more relational model of higher education,” we must return to the way things were done, and he then proceeds to cherry-pick different speech and writing assignments all the way back to the 1300s. In the process, I think he indirectly describes a lot of the pedagogy common to small discussion classes like first year writing: requiring students to meet during office hours, entering into “Socratic dialogue or simple Q&A” with the class, and so forth.
“There is the problem of scale,” with old techniques like oral exams, Shirky admits. “With some lecture classes in the hundreds of students, in-class conversation is a nonstarter.” Well, wait a minute: maybe the past practices we need to return to are smaller classes. Perhaps one of the reasons why I am not that worried about AI cheating is that I feel like I actually do most of these things in the classes I teach now. My students end up doing a lot of writing— discussion posts to readings, scaffolded essays part of the research project, and drafts of work in progress— along with plenty of discussing as well.
So what if every class were no more than 25 students? That wouldn’t be logistically possible, and it wouldn’t be a complete solution to AI cheating either, of course. But it’s a start, and we’ve also known for a very long time that lecturing is also a terrible pedagogy.
I will say this: both end on a vaguely positive note, even if their optimism about the future does not strike me as particularly realistic. Good takes a lot of pleasure in this return to the past, connecting us back to Plato and education as “not a process of pouring knowledge into an empty soul, but as a ‘turning around’ of the soul in the direction of beauty and truth.” She sure seems to think that those blue-books and handwriting can accomplish a lot!
And after spending the rest of his op-ed saying there’s nothing to be done about AI except return to “technology free” classrooms, Shirky ends by predicting higher education will adapt. “Despite frequent pronouncements that college is doomed because students can now get an education from free online courses or TV or radio or the printing press, those revolutions never flattened us. Nor will A.I.” We’ll see. I want to believe Shirky is right, but….