Classroom Technology Was a Mistake

Hopes that AI will improve higher ed need to reckon with the dashed hopes of the past

A photo of a horrible policy mistake. Photo by Gustavo Fring: https://www.pexels.com/photo/boys-working-on-computers-in-classroom-5621935/

I think my micro-generation (Xennials, or the Oregon Trail generation) was the first for which “technology in the classroom” was a guiding principle of our educators. Now, there’s always some wiseass scholar in the audience who wants to say that “technology has always been in the classroom” and they want to talk about how pens or pencils or flint arrows are “technology”, but we all know that what we mean are computers—the most powerful general-purpose tool of the past one hundred years. At the very least, the idea of bringing computers—physically—into classrooms or classroom-adjacent spaces (the venerable computer labs, an intermediate stage of computational integration now lost to the ages) cannot be much older than I am, because the idea of a computer small enough to be used by five-year-olds is not much older than I am—we are, now, the same age, even if when I was a kindergartner those extra five years seemed like an eternity.

I cannot remember educational spaces without computers, even though in many cases they must have been installed just a couple of years, or maybe even a few months, before I toddled onto the scene. I never really learned to use a card catalog, because my library’s catalog had been digitized; I recall distinctly being in a computer lab with a cluster of PCs—or, more technically, Commodore 64s—as early as first grade, and I remember looking with envy at the students in that lab when I was a kindergartner.

Now, computers are ubiquitous in the classroom. It would be strange to work in a classroom with none. It is not uncommon now for a student in one of my classes to have up to four computers with them in class—a laptop, of course, but also a phone, a tablet, and a smartwatch.

How fantastic—in the old sense—this casual accumulation of compute is. There was a time when smartwatches were purely science-fictional, even pulp-fictional; Dick Tracy had one in the movie I saw (too young) in theaters. Captain Picard and Penny from Inspector Gadget had tablets—one in the far future, the other in a context in which policemen had helicopter rotors in their heads. And we all know by now that our personal computers are so much more powerful than the Commodore 64 that if airplanes had gotten faster at the same rate they’d now be flying shuttle routes to Alpha Centauri, or something.

For all purposes, then, computers in education and I have come of age together. Have they delivered on the gleaming promises that forward-looking teachers and officials subscribed to in the 1980s?

No, absolutely not, and they’re probably mostly a disaster. We’d likely be better off without them most of the time. To the extent they are good—for data analysis classes or rapid consultation of research databases—they are used surprisingly rarely. Far more often they’re used as distraction engines or counterfeiting machines.

They have not even made much of the daily business of administering an educational enterprise better. To be sure, there’s something to be said for not having to work with Scantron bubble sheets or paper gradebooks anymore, especially for those of us cursed with illegible handwriting. As someone who once lost a stack of essays to be graded in the Before Time, I can also appreciate that we can save essays to the hard drive or to the cloud; typewriters are romantically nostalgic for about five minutes, but the first time you realize that cut and paste used to be a descriptive phrase and not a metaphor you’ll be reaching for the MacBook instead.

Yet the actual technology we actually use is mostly subpar in terms of its effectiveness. I’m writing this as a timed exercise—20 minutes on the timer, no more—in an effort to purge myself of the frustrations of working with Canvas. Canvas, as most of you doubtless know, is the LMS (Learning Management System) used by my university and others; there’s also Blackboard, Moodle, and many more in the sector.

The promise of Canvas is that it will bring all of the fiddly back-office parts of running a class into the cloud seamlessly and easily: gradebooks as spreadsheets, readings as PDFs, exams as Web pages rather than bluebooks. In reality, the software is — probably inescapbly — incredibly unwieldy. One example that has driven me to distraction today: there is no easy way (at least in our instance of Canvas) to set up an assignment that repeats daily. In the old days, when syllabi were printed out (and ran to four or five pages), setting up a recurring assignment required typing “Every Wednesday, hand me a two-page paper.” Now, it requires at least—at least—an hour of fiddling and re-fiddling and debugging, because the software isn’t set up to do this sort of task—an absolutely minimal requirement—a push-button exercise.

The tools, in other words, don’t even make the bad parts better. Often, they’re worse. Not a few professors opt out of using them, which is truly damning. Nobody chooses to hand-copy assignments; they just use the photocopier. But a number of professors choose—illicitly—to run their courses from personal Web sites or Dropbox accounts or anything other than the LMS.

So why use them? Well, we are trapped by our tools at time, but we are also trapped by the expectations of students and the policies of our employers. The latter is not a small obstacle, by the way; if you post links to databases with readings through the official LMS, you are not infringing copyright, but if you host a reading on your personal Web site, you definitely are. And if you let personal information leak from your personal site, you’re guilty of transgressions twice over (and, I must say, justly so—the data isn’t the instructor’s and no instructor has the capability to actually secure their site in an era in which university records are the targets of, say, political hackers going after college admissions essays).

The former, however, is also important; we are now at least fifteen years into a transition in which LMSes have been creeping into high schools, elementary schools, and even nurseries. Students and their parents alike expect that this is how it will be. Doing anything else would require retraining the students’ expectations while looking like a weirdo and inheriting all the risks of a paper world (like, you know, losing stacks of exams or having homework devoured by ravenous canines).

That digitalization looks modern and efficient but it often is anything but. It takes me hours longer to create a course shell than it used to and the results are not appreciably better; indeed, because of shortcomings in the tools, I find myself frequently battling between adapting my course to the machine or bashing the machine into doing what I want.

It does not help that LMSes are walled gardens, and weedy ones at that. Much of what I want to do could be done easily in a vanilla setup but everything that’s been added for commercial or regulatory purposes gets in the way of the simple business of sharing assignments and readings. And, note, that none of this is doing anything but replacing, at great time and increasing expense, what used to be done without paying licensing fees or technicians to assist. Somehow Feynman didn’t need this (or if that’s too strong pick someone who was an actually good undergrad instructor).

The story repeats in many scenarios and applications. For every minor friction that technology has removed, a giant weight of superfluous expectations has been added. If the net impact is still positive, it is much less so than the boosters promised—so much less so that one wonders whether we would have gone all-in on this. And I also find myself wondering if at least some of the lack of grit that many of the younger generations display when confronted with an obstacle has to do with being raised in an environment in which everything just works—where you never have to slap the side of a TV to fix the image, or when you never have to parse “Abort/retry/fail” when trying to load a program. To the extent that computers do make everything “just work”, they do so by bounding the ability of users to ask for anything — limiting their requests, and their imaginations, to what’s easiest to deliver. But that is far from what’s possible and often remarkably distant from what’s desirable in any given case.

I am not an “opponent” of AI; I don’t even consider myself a “skeptic” of LLMs. I think that they have many more use cases than the most overworn cliches (“they’re just autocorrect!”) admit. But I also wonder whether moving wholeheartedly into adopting LLMs in the classroom is a triumph of hope over experience. Is it really likely that the LLM will be able to contribute more than, say, Microsoft Encarta CD-ROMs did to a fourth-grade classroom? Or is it more likely that introducing a giant pipe into the oeuvre of human kind is at least as likely to convey sewage as anything potable? If we can’t even get an LMS vendor to deliver “repeat this task every class day” as a normal setting — if there’s not even a feature to auto-populate a syllabus with every day a class will meet in the LMS; you know, the sort of boring database management computers are unparalleled at — then why should we expect that widespread LLM adoption will end up as anything less than an edutechnical innovation that impresses the rubes and fails to move the needle on actual outcomes?

I do think there are paths available to a better outcome, or sets of outcomes, but like all such paths they mostly involve tradeoffs from the bright shiny sales pitch. Do you really think you get all the advantages but with no extra costs? Would LLMs look as appealing if we described them as a useful tool that requires (as they do!) extra human involvement and mentoring to make use of their full potential—and that the time measured to reach those goals would be measured in years of careful engagement?

The example of computerized classrooms suggests that thoughtfulness will not be the distinguishing characteristic of what happens. It is even unlikely that those “at the coalface” will be consulted or valued as policies are made; the managerial-consultant class of higher education is quite happy to tell line faculty what the line faculty should think about these exciting new opportunities. If one is hired to be a consultant on these issues, after all, one is selling the ability to make LLMs a part of the classroom, not to judge whether they should be kept, like predators, a thousand feet away from students at all times.

At root, of course, is that all computing technology since the 1980s has been sold as a way to bring tools into the curriculum but those efforts have mostly brought toys into the classroom. (That my generation is known, in some circles, after a computer game, The Oregon Trail, which famously failed to teach any of us anything about the Oregon Trail itself is a grim irony.) As tools, computers are godsends; I am much more happy with LLMs in my mode as a researcher than as a teacher. As a classroom manager, I find myself pining for the days when I just needed an Excel spreadsheet and a stapler to manage my students. And as an instructor, I wonder whether there isn’t something to be said not only for bluebooks and pens but for slide rules and chalk—analogue tools to develop organic intelligence.