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….

I’m kind of surprised, but I am still coming across essays and Substack posts and such where teachers/professors are freaking out about AI. ChatGPT came out in November 2024, more than two and a half years ago. I would have thought folks would have moved on from these “writing assignments are dead” kinds of pieces by now, but no–throw a brick out a window and you’ll hit one. Here’s a good recent example: “The Death of the Student Essay– and the Future of Cognition” by Brian Klaas. The title is the gist of it– I’ll come back to Klaas’ essay later.

It’s not that these “the death of the assigned paper and now I’m going to make my students chisel everything into stone” eulogies are entirely wrong. As I’ve been saying for a few years now, AI means teachers who used to merely assign writing with no attention to process can’t do that anymore. AI means teachers need to adjust their approach to education. It doesn’t mean that all of a sudden everyone will stop learning.

And before I go any further, I kind of think what I’m writing about here is Captain Obvious wisdom, but here it goes:

Here’s what I mean:

Learning is about gaining knowledge and skills, and humans do this in lots of different ways— play, practice, observation, experiences, trial and error. We learn things from others and the world around us, and while learning is often frustrating, I think learning is pleasurable and fulfilling. All of us start learning right after we’re born— how to get attention, to crawl, to roll, to walk, etc.— through help from our parents of course, but also on our own.

Some things we learn through exposure to the world around us; for example, speech. Of course, parents and others around babies try to help the process along (“say da-da!”), but mostly, babies and toddlers learn how to speak by picking up on how the humans around them are speaking. And as anyone who has parented or spent time around a chatty pre-schooler knows, sometimes it can be challenging to get them to stop talking.

On the other hand, some things we need to be taught how to do by others— not necessarily teachers per se, but other people who know how to do whatever it is we’re trying to learn. Reading and writing are good examples of this, which is one of the ways literacy is different from speech (or, as Walter Ong might have put it, orality). This is one of the reasons why, up until a few hundred years ago, the vast majority of people were illiterate.

Except Tarzan. This is a bit a of tangent, but bear with me:

Edgar Rice Burrough’s famous novel Tarzan of the Apes is an extraordinarily interesting, odd, offensive novel, and most of the adaptations of the book gloss over its over-the-top fantasy and weirdness. At the beginning of the book, Tarzan’s parents are put ashore in Africa after a mutiny on their ship, and his father builds a cabin stocked with the goods Tarzan’s parents were traveling with, including a lot of books. The parents are killed by “apes” (which are somehow different than gorillas, but that’s a different story) and the baby that becomes Tarzan is raised by them.

When he is around 10, Tarzan stumbles across the cabin with its books, and, long story short, he teaches himself to read. He does this by staring at the the marks on the pages of a children’s book, letters that looked like little bugs next to a picture of a strange ape that looked like him, and he figured out those little bugs were b-o-y. ”And so he progressed very, very slowly, for it was a hard and laborious task which he had set himself without knowing it—a task which might seem to you or me impossible—learning to read without having the
slightest knowledge of letters or written language, or the faintest idea that such things existed.” Basically, Burroughs is saying “yeah, I know, I know, but just go with it.”

In contrast, education is a technology. To quote from my book, education is the “formal schooling apparatus that enables the delivery of various kinds of evaluations, certificates, and degrees through a recognized, organized, and hierarchical bureaucracy. It’s a technology characterized by specific roles for participants (e.g., students, teachers, professors, principals, deans) and where students are generally divided into groups based on both age and ability.” This is an argument I belabor in some detail— you can read more about it here with the right JSTOR access— but I’m sure anyone reading this has had first-hand experience with what I’m talking about.

Learning and education are a Venn diagram: when schooling goes well, education facilitates learning, and successful learners are rewarded by their educational experiences with degrees and certifications. But sometimes schooling does not go well. For whatever reason, some students, especially in courses like first-year writing, just do not want to be there. That was the case for me in a lot of high school and college classes. Sometimes, it was because of bad teaching, but more often than not, it was my lack of interest in the subject, or the fact that it was a subject I was (and am still) not very good at– anything having to do with math or foreign languages, for example. Whatever the reason though, I knew I had to push through and do the course in order to move on toward finishing the degree.¹

Everyone involved in education gets frustrated by the bureaucracies and rules of it, especially when the system that is education gets in the way of learning. For example, even professors in business colleges are annoyed by students who are not there to learn anything but to just get the credential and the job. Students are often annoyed at their professors who don’t seem to know how to help them learn because they are just so bad, and everyone is annoyed with all of the other curricular hoops, paperwork, and constant grading. And that’s because learning is the fun part, and the important part!

But here’s the thing: the occupational, monetary, class, and cultural values of academic credentials– that is, the degree as a commodity– are only possible with the technology of education. It is why students and their families (our “customers”) are willing to pay universities so much money. As I wrote in my book, “Students would probably not enroll in courses or at universities where they didn’t feel they were learning anything, but they certainly would not pay for those courses if there was no credit toward a degree associated with them.”

Educators, and I like to think most students as well, are attracted to the university because they enjoy learning and place a high value on learning for the sake of learning: that is, the humanness of it all. But look, I don’t know anyone who is a teacher or a professor who does this work just for the love of it. This is a job, and if I didn’t get paid, I wouldn’t be doing it. Besides, there is a lot of value in education’s certifications and degrees in all of our day-to-day lives. I find it reassuring that the engineers who designed the car I drive (not to mention the roads and bridges I drive on) have degrees that certify a level of expertise. I am glad my dentist went to dental school, that my doctor went to medical school, and so on.

So, to circle back to how this connects with AI in general and with Brian Klaas’ essay in particular: I think the vast majority of the “AI and the end of student writing” essays I have read (including this one) are incorrect in at least two ways. The first way, which I have been writing about for a while now and which I mentioned at the beginning of this post, is about the distinction between assigning writing as a product and teaching writing as a process. Like most teachers, Klaas does not seem to have a series of assignments, peer reviews, opportunities to revise, etc.; he’s assigning a term paper and hoping students write something that demonstrates they understood the content of the class. Klaas writes “Previously,” meaning before AI, “there was a tight coupling between essay quality and underlying knowledge assembled with careful intelligence. The end goal (the final draft) was a good proxy for the actual point of the exercise (evaluating critical thinking). That’s no longer true.” By quality, I think Klaas means grammatical correctness, and I don’t think that has ever been the primary indicator of a student’s critical thinking. Yes, the students who write the best essays also tend to write in grammatically correct prose, but that’s a pretty low bar. And don’t even get me started on the complexities scholars in my field could unpack in Klaas’ claim about the “coupling” between “quality” and “intelligence.”

Klaas also doesn’t seem that interested in doing the extra work of teaching writing either. He writes:

More than once, a student quite clearly used ChatGPT, but to try to cover their tracks, they peppered citations for course readings—completely at random—throughout the text. For example, after a claim about an event in 2024 in Bangladesh, there was a citation for a book written ten years earlier—about the Arab Spring. “Rather impressive time machine they must have had,” I commented.

After a career working to develop expertise, countless hours teaching, and my best attempts to instill a love of learning in young minds, I had been reduced to the citation police.

I’m sure Klaas is correct and this student was cheating, but I’ve got some bad news for him: if you want students to use proper citation style, you have to teach it. And, as I’ve written about before, teaching citation is even more important with AI for a variety of reasons, including the fact that AI makes up citations like this all the time.

But again, Klaas doesn’t want to teach writing anyway; “Next year, my courses will be assessed with in-person exams.” Well, if Klaas was assigning writing so students could write essays that are like answers to questions in an exam, maybe he should have just given an exam in the first place.

This leads me back to my Captain Obvious Observation: learning and education are not the same thing. Yes, any of us can use AI as a crutch to skip our innate needs and desires to learn, but AI’s real impact is how it disrupts the technologies and apparatuses of education. Klaas says as much, ultimately. He points out that AI probably means “universities will need to find ways to certify that grades are the byproduct of carefully designed systems to ensure that assessments were produced by students.” And in passing, he writes “We must not fall into the trap of mistaking the outputs of writing (which are increasingly substitutable through technology) from the value of the cognitive process of writing (which hones mental development and cannot be substituted by a machine).”

Exactly. And I think we know how to do that.

First, we have to teach students about AI, and that’s especially true if we don’t want them to use it. For example, had Klaas explained to his students that AI makes up citations all the time, they might not have tried to cheat like that in the first place. It’s not enough to just say “don’t use it.”

Second, we need to lean more into learning, and we need to be more obvious in explaining to our students why this is important. Teachers need to do a better job of explaining to students and ourselves why we ask students to do things like write essays in the first place. It’s not just so teachers have something to assess as evidence of what grade that student deserves. That’s education. Rather, we have students write essays (or write code, do math problems, conduct mock experiments, etc.) because we’re hoping they might learn something.

Third, we need to change how we teach in ways that discourage relying too much on AI and encourage students to do the learning themselves. Unfortunately, this is a lot of work, and I think this is actually what Klaas and others lamenting the “death” of student writing are really complaining about. The “write a paper about such and such” assignments faculty have been relying on forever won’t work anymore. Though maybe that assignment you thought worked well before AI actually wasn’t that effective either?

1. “Moving on” did not necessarily mean finishing the course– I dropped several as an undergraduate to avoid a D or an F. Also, I was lucky and unlucky as an undergraduate when it came to my two weakest school subjects. For my degree in English back in the 1980s, I did not have to take any math courses at all. However, I was required to have four semesters of a foreign language. If I had had to take the math class that my EMU English majors have to take as part of general education, I’m not sure I would have made it. On the other hand, EMU students do not have to take a foreign language. I studied German, and I was terrible at it, which is why it took me about seven tries (including a summer school class) to pass the four semesters I needed. ↩︎