Artificial Mediocrity: The Hazard of AI in Education

AI-generated text provokes soaring hopes for limitless potential. Programs like Chat GPT and GrammarlyGO seem like wonderworkers. They “empower,” “assist,” and “inspire” their users. But after the first full semester of ChatGPT’s ubiquitous appearance in American classrooms, there is good reason to think that, far from helping students, chatbots imperil the very possibility of serious education. Chatbots replace disciplined learning with unthinking suggestibility and encourage students to avoid exercising practical judgment. Teachers must convey the hard reality that using a chatbot to skip the stages of an assignment that require organizing one’s own thoughts and research is not just dishonest, it is stultifying. It precludes excellence, and it encourages mediocrity.

Transforming Education

When it comes to education, excitement about chatbots is due to more than just giddiness about a gadget that passes the Turing test—the famous benchmark of AI that may not be as significant as previously thought. This accomplishment dazzles some onlookers into believing that ChatGPT is on its way to becoming sentient. However, true believers in educational chatbots are hoping for a radical shift in education itself. In fact, they perceive chatbot assistants as the development that will finally modernize teaching and learning.

Technology is supposed to improve the efficiency of education by helping students avoid difficulties that would otherwise slow them down. Chatbots are said to improve the educational experience for students by making suggestions that lead to more personalized writing, helping them produce “your best work faster,” as Grammarly claims about its chatbot.

The most eager innovators argue that ChatGPT is killing backward inefficiencies in the learning process. In the past, teachers relayed information for students to summarize and repeat. ChatGPT purportedly enables a more sophisticated approach that produces “knowledge transformers, rather than knowledge dispensers.” It does this by helping students skip the harder initial stages of assignments, where they tend to get stuck trying to condense facts and organize their thinking. An article for the MIT Technology Review approvingly describes a teacher who now asks her students to annotate AI-generated arguments instead of drafting their own, letting them focus on the critical phase of editing instead of the upfront work of synthesizing and processing information on their own.

All these hopes rest on the assumption that producing a first draft or organizing one’s research is like drafting a routine email. But is it the case that real education doesn’t start until things “get moving?” Are the purported obstacles that chatbots help us circumvent just bumps in the road, or might the difficulties we encounter be a vital part of what is required for us to learn in a serious, lasting way?

AI and Comfort Culture

Infatuation with educational AI is symptomatic of what Greg Lukianoff and Jonathan Haidt dub “the coddling of the American mind,” the movement in higher education to protect students from psychological discomfort. In contrast with older traditions that ask uncomfortable questions and stress the virtue of conformity with truth, chatbots cater to the illusion that learning, writing, and researching should be easy. As with other educational innovations, the aim is to improve performance by streamlining the process—that is, by making it less challenging.

But serious education is provocative. The effective teacher pits the radically curious side of a student’s mind against the indolent and complacent side. The goal of excellent teaching, as Jacques Barzun put it in Teacher in America, is to turn a student, “by nature a little copycat, into an independent, self-propelling creature, who cannot merely learn but study—that is, work as his own boss to the limit of his powers.” An AI “co-creator” designed to make writing delightful by tailoring homework to fit a preferred style and mood prevents the sort of independence that serious education cultivates.

One need not appeal to the authority of “the old ways” to refute the aspiration for frictionless learning. The psychologist Robert Bjork confirmed the need for “desirable difficulties” in education. Bjork found that “conditions of learning that make performance improve rapidly often fail to support long-term retention and transfer, whereas conditions that create challenges and slow the rate of apparent learning often optimize long-term retention and transfer.” In other words, practices that seem pragmatic because they raise grades and garner student approval do not work. Loose group assignments, eleventh-hour review sessions, and non-comprehensive assessments merely pacify. By contrast, Bjork found that the time-honored ways of testing frequently, penalizing cramming, and assessing comprehensively (all practices that students tend to loathe) are precisely what make lessons stick. A bit of friction—what educational technology attempts to overcome—is exactly what the mind needs.

Practical Wisdom and the Art of Writing

AI text generation provides the student with what he wants in a seemingly rational way, but without involving the student’s own rationality. There is no need to think over what one is saying when using a generative AI assistant. Rather, the writing process will be more efficient to the extent that the student does not think it over. To personalize writing, the student need not be personally involved in discovering the word or phrase to be used. The hope to attain frictionless learning through AI assistance is predicated on the belief that we can circumvent sound judgment—a quality that is by no means distributed equally.

But discovery is the essence of composition. Aristotle and Cicero define rhetoric as the art of discovering (or inventing) the means to persuade. As a medium of rhetoric, writing is heuristic (a word we inherit from the Greek eureka, “I have discovered!”). It involves practical judgment, not mathematical certainty, to discover how to communicate a coherent argument in a manner that befits one’s reader. Thus, writing is closer to carpentry than to computer programming. It requires a capacity for judgment—for what the ancients called “practical wisdom”—that AI language models lack. (ChatGPT confessed to me that an AI language generator will always require “human critical thinking” to calibrate its responses.) As one of my best writing teachers would say, “The point of a sentence is to pass sentence.” If you are not acting as a prudent judge over hard cases, you simply are not writing. More than that, you simply are not thinking.

Appealing to Excellence

Teachers must continue to discourage the use of AI programs to produce assignments (in whole or in part). To this end, AI-detection programs like Turnitin are beneficial, as are handwritten exams, and oral examinations are uniquely suited for evaluating students’ knowledge outside typed assignments.

Yet we need much more than mere policing: teachers must impress on their promising students that the use of AI dulls and mediocritizes the mind. In a letter attributed to St. Thomas Aquinas, one finds these conjoined maxims: “ea quae legis et audis, fac ut intelligas; de dubiis te certifica,” loosely translated as, “Make certain that you understand what you read and hear; resolve doubts for yourself.” Moral and intellectual excellence intersect on this point: There is no substitute for responsibility. One either undergoes the discipline necessary for reaching a serious standard or one outsources one’s mind (and soul) to an enticing “assistant” that satisfies the visceral desire for comfort.

Teachers should also echo (and apply) the hard truth of education that C. S. Lewis articulates so clearly: “Virtue is not democratic; she is achieved by those who pursue her more hotly than most men. Truth is not democratic; she demands special talents and special industry in those to whom she gives her favours.” Those who take care to read, listen, question, and write well are invested in a praiseworthy undertaking. The teacher should honor them for choosing this harder path when most of their peers have settled for Google queries. Many students may not appreciate the importance of applying themselves rather than using AI, but we must encourage those who do. We should fortify promising students with the assurance that excellence in education is worth pursuing but requires taking a hard road.

Vice and Technology

One could argue that chatbots are not inherently an obstacle to serious education. What if a student simply uses a program like GrammarlyGO as a tool without letting it prevent independent thought? Unfortunately, the problem is with the use of chatbots, not just the abuse, as is often claimed in popular media. Shilo Brooks recently wrote on the notion of “natural technology” in Antón Barba-Kay’s A Web of Our Own Making, explaining that such technology is “so intuitive and effortless to use that it does not feel to humans like they are using something external to themselves.” Chatbots are designed to facilitate frictionless thinking and writing. No one would ever suspect spellcheck of being able to satisfy such promises (as we all know, it barely fulfills the task for which it is made).

A student who sets out to use the GrammarlyGO chatbot may intend to check for spelling and punctuation errors, but the program is designed for more than this, and its offers are alluring. Why not explore what changes in syntax would give my writing a “warmer” tone? Moreover, why not use the chatbot for suggestions that will help me conquer writer’s block? Students I know who began using chatbots for reasons like these slipped into producing whole drafts with the help of AI. The initial promise that learning is better when chatbots remove obstacles gradually leads to the “abuse” of a tool that replaces (and debases) the minds of students. Students would be better off taking the time to consult Strunk and White’s Elements of Style and Fowler’s Modern English Usage than cutting corners with the help of a seemingly efficient chatbot assistant.

The alluring combination of vice and technology that AI tools feed is attractive in mainstream academia and across various industries. But it is discipline and a willingness to think for oneself that conduce to excellence. Surely, both the old-fashioned teacher and the old-fashioned journeyman would recognize this principle. As the cowboy poet Red Steagall mused, “Now we could let it go like this / And take the easy route. / But doin’ things the easy way / Ain’t what it’s all about.”

Image by Firn and licensed via Adobe Stock. Image resized.