Because I have snow day (!) and some actual free time on my hands, I’d like to offer up this slightly less long-winded follow up to my previous post, At this point, I’ve spent many hours reading various articles, blog posts, and discussion forums about the effects of AI in the humanities classroom, and one recurring theme I’ve noticed is the newfound appreciation of mediocre work that has clearly been produced by actual human students themselves rather than spat ready made out of a machine. Such students–who would have thought?–are apparently the newfound heroes of the AI age.
It is not exactly hard to imagine why this is the case. For a professor who has no choice but to treat piles of AI-generated smarm as if they are in fact students’ own work, an awkwardly expressed piece of (post-)adolescent writing that shows some level of engagement with the material at hand can understandably seem like a ray of light beaming down from the heavens.
I understand this relief, and the ensuing temptation to celebrate it, but I also think it is important not to take this tendency too far.
I hesitate to use words like “standards” and “expectations” here because they have been reduced to such clichés; however, it is difficult to discuss this matter without invoking them to some extent. What concerns me, fundamentally, is that human-produced work itself, however trite or mediocre or poorly expressed, will increasingly come to be perceived as sufficient in and of itself—that is, there will be no need to encourage students to aim for anything higher, competence and polish being exclusively associated with bot-produced prose (derived, of course, from the unauthorized and uncompensated use of countless professional writers’ work, but let’s leave that aside for the moment).
Witness, for instance, the confusion over whether real people use em dashes in their writing, and whether the appearance of this form of punctuation is a sure sign of mechanical generation. I hate to break it to anyone arguing favor of the latter, but yes, professional authors do use en dashes all the time—that’s where AI got them from. The fact that people are wigged out by this punctuation suggests that they haven’t been exposed to enough adult writing to even be able to recognize what it looks like.
Look, learning to write well is a longterm, messy process, and much student writing is by definition unrefined. It takes a lot of practice, and a lot of reading, and a lot of carefully directed feedback, combined with a willingness to implement that feedback, to get good at it. And to state what should be obvious but now needs to be repeated continually, if students are using ChatPGT to compose their papers, they are not learning how to write. That’s kind of like trying to get in shape by going to the gym and sitting on the bench while staring at your phone (which, for the record, is something I see on pretty much a daily basis).
Resisting the lure of auto-generation and exhibiting a willingness to do the hard work of getting one’s thoughts out onto the page is commendable; however, it does not therefore follow that things should stop there. I have learned not underestimate the American education system’s ability to lower expectations, but seriously, where exactly is the limit for people in higher education who claim to value that nebulous faculty known as critical thinking? Will the Modern Language Association, for example, follow the National Council of Teachers of English in declaring that “media literacy” should take primacy over essay writing? Will humanities classrooms be turned into podcasting centers?
The fact that AI can spit out slick, hyper-organized, superficially impressive prose does not imply that students should not learn to write clearly or organize their thoughts or use academic vocabulary in the service of seeming somehow more human. To argue otherwise is to imply that clarity and organization are the province of machines, whereas human beings should be satisfied with nothing more than stream-of-consciousness ramblings. (I read many of these as a tutor; trust me, I know the type of writing of which I speak.)
This is an astoundingly diminished expectation, not to mention an incredibly depressing—and inaccurate—view of human potential. People, after all, wrote the works that LLM responses are essentially pastiches of; people came up with the ideas and developed those types of structures in the first place. The machines have merely learned to simulate them statistically.
The ability to experience emotions may be a mark of humanity, but so is the ability to apply one’s critical capacities. Giving too much credit to bot-produced work implies that students should conflate text is merely superficially impressive with writing that exhibits genuine sophistication, and that it is not worth their time and effort to bother with learning to differentiate between the two.
Crucially, there is no shortcut for acquiring the ability to make this type of distinction. You have to read a lot of high-quality writing to develop a sense of what a knowledgeable authorial voice feels like, and you need to know a fair amount about whatever you’re reading about to be able to identify when things don’t make sense, or are based on straw-man arguments, or are just plain wrong. At the same time, you cannot become so willfully skeptical that you reject well-supported arguments because that’s just propaganda (as one of my students was fond of pronouncing dismissively) and what do those so-called experts know anyway? Learning to manage that tension is one of the main goals of education.
In terms of writing, learning to control your language, to shape it and bend it to your needs in order to refine your thoughts, and use it to arrive at a well-considered conclusion—that is learning to think critically. In a media and entertainment ecosystem based on quick dopamine jolts and split-second reactions, the ability to slow down, think things through, and attempt to understand how they actually function is, I would argue, far more important in the long term than the kind of “relevance” based on reducing classwork to the level of a social media feed. I would also argue that this type of work can be also be, dare I say it, a form of pleasure. Otherwise, you remain at the mercy of the algorithms, perpetually buffered this way and that, like Dante’s Paolo and Francesca in Hell, or caught like a figure in a snow globe inundated by an endless whirl of stuff, unable to perceive or even imagine a path forward.