Will Artificial Intelligence Make Us Dumber?

Since ChatGPT¹ made a splash in 2022, the prevailing sentiment of artificial intelligence in higher ed. has been one of doom: the more students use A.I. to write, the worse their natural ability to write becomes. Articles like “ChatGPT’s Corrosion of the Classroom: AI denies student writers the chance to connect” and “What we lose when machines do the writing” question the impact of A.I. when used in educational settings.

As a college professor, I too shared the fear that using A.I. will compromise students’ natural ability to communicate. But now that we’re three years into the Age of A.I., and considering how prevalent the use of A.I. has become, my fear about A.I.’s negative impact has grown. My original concern was that college students would plagiarize their way through college and wouldn’t develop the necessary and fundamental writing skills any educated graduate should possess. My fear was that the “average student” — representative of the student population at the community college where I teach — would be using A.I. to the point that they’d graduate without the natural ability to write at the college level. This fear was shared by Professor Jeff Schatten in his 2022 article “Will Artificial Intelligence Kill College Writing?”.

Yet, I now believe the most important casualty of A.I. is not writing. It is reading and thinking.

In one of my courses, I assign Mary Shelley’s Frankenstein along with a paper about the novel’s Romantic sensibilities — a boilerplate assignment for the humanities at the 1000/2000 level. Last semester, I suspected that at least half of the submissions were written by A.I.² When I looked into my course analytics, I saw a surprising data point: about a third of the students in the class didn’t click on the link to read the book. One conclusion we could draw from the data is that students are reading the material elsewhere, not using the provided link. Another might be that students are going straight to A.I. software for the answers to assignments they haven’t bothered reading.

The probability of the second scenario being the correct one is high. In another assignment on seventeenth-century philosopher Pierre Bayle, a third of students discuss examples from the text that don’t actually appear in the supplied text. I suspect this sort of thing will decrease in the future. Not because students are cheating less, but because A.I. is getting “smarter” and students are getting better at using it. Generating an essay from a lengthy, complicated reading and a set of instructions is no harder than copying and pasting them into an A.I. prompt.

That students are using A.I. without bothering to read mirrors how the public interacts with AI. Take, for instance, Google’s recent “A.I. Overview” feature, which presents users with an A.I.-generated response to a search query. Instead of users browsing the web and potentially reading a variety of websites and resources, Google generates a summary. A.I. software itself is discouraging reading and thinking in the name of saving time and hassle. Google says “our custom Gemini model can take the legwork out of searching.” People who might not even know they’re using A.I. are being discouraged from researching, reading, and thinking. This is not just speculation. Bloggers, writers, and others who make a living publishing on the Internet have been complaining that website visits and search traffic has been negatively impacted by A.I. tools like “AI Overview.” This point has been all but confirmed by Big Tech. For many, analytic data measuring web traffic and ad revenue all but confirm the suspicion: people are visiting websites less (and thus reading less and earning less ad revenue for online publishers) in the Age of A.I.

This idea was explored in a recent New Yorker article, “Will the Humanities Survive Artificial Intelligence?” The author, Princeton professor D. Graham Burnett, seems to agree with me. But he seems to shrug off the fact that students will be reading fewer rigorous texts. He also envisions a future wherein there isn’t even a need to write them. Burnett states:

I can construct the “book” I want in real time—responsive to my questions, customized to my focus, tuned to the spirit of my inquiry. And the astonishing part is this: the making of books such as those on my shelves, each the labor of years or decades, is quickly becoming a matter of well-designed prompts. The question is no longer whether we can write such books; they can be written endlessly, for us. The question is, do we want to read them? … Within five years, it will make little sense for scholars of history to keep producing monographs in the traditional mold—nobody will read them, and systems such as these will be able to generate them, endlessly, at the push of a button.

Like a game of hot-and-cold, Burnett gets close to the problem, but he never succeeds in grasping it. Forget about dense scholarly monographs and academic tracts; nobody reads those anyway. A.I.’s ability to ingest, analyze, and generate is a risk to reading in general.

Getting students to read is, like plagiarism, a longstanding problem. One pre-A.I. study found that 65% of students surveyed reported not buying textbooks (costs were reported as the predominant reason). Yet it seems that, even when the reading material is free as it is in my class, the use of A.I. is making the problem worse. This is on top of the decline in reading brought on by smartphones, the attention economy, and algorithmically mediated feeds.

In a recent article in the Financial Times, journalist John Burn-Murdoch charts a downward trend in reading among teenagers since the 1980s. Now, nearly 50% of teenagers report “hardly ever” reading and less than 15% report reading “almost every day.” The author correlates this data with a few other data sets, ones that chart a similar decline in reasoning, problem-solving, thinking/concentration, and learning. He concludes that “we appear to be looking less at the decline of reading per se, and more at a broader erosion in human capacity for mental focus and application.” This is a result of a “shift in our relationship with information.” What the author does not consider, however, is how the use of A.I. changes our relationship with information: it sifts through a vast library of information we would otherwise read and process before presenting to us a concise answer to the question we ask of it.

Cognitive Death Spiral

Georgetown Professor and author Cal Newport notes in a recent podcast episode that the ecosystem of smartphones, social media apps, algorithmically mediated feeds, and our new relationship with information have introduced a “cognitive death spiral.” He concludes that this ecosystem not only damages our ability to concentrate and sustain focus, but it keeps us from engaging in activities that make us smarter. Like reading. What results is a feedback loop of behaviors that are not conducive to markers of intelligence like concentration and critical thinking.

If my experience in the classroom is representative of a larger trend in society, that the use of A.I. is precipitating a decline in reading and thinking, then its use creates a similar “cognitive death spiral.” A.I. will chip away at our natural ability to communicate, and it will ensure we don’t enjoy the benefits commonly associated with reading like improving memory, creating new neural pathways, slowing cognitive decline, improving focus and concentration, developing empathy, and sharpening our critical thinking skills. If reliance on A.I. will increase, our cognitive decline will spiral out of control.

Complaints about the decline in reading and writing are as old as the written word. One of the translated Sumerian clay tablets recorded complaints about the declining writing ability among the youth of Mesopotamia.³ Yet it seems the complaints increased with the arrival of electronic culture. Sven Birkerts lamented the decline in reading culture in his (must read) 1994 book The Gutenberg Elegies: The Fate of Reading in an Electronic Age. The complaints have continued. In his 2024 article “The Loss of Things I Took for Granted,” professor Adam Kotsko reports that his students cannot read effectively, adding that “students are intimidated by anything over 10 pages and seem to walk away from readings of as little as 20 pages with no real understanding.” More recently, one Ivy-League philosophy professor wrote on Substack that “most of our students are functionally illiterate.” Rose Horowitch reported in The Atlantic that students are having an increasingly difficult time reading books cover-to-cover even in Ivy League schools, a problem brought on by electronic devices and the lack of reading in middle and high schools. Horowitch writes: “In 1976, about 40 percent of high-school seniors said they had read at least six books for fun in the previous year, compared with 11.5 percent who hadn’t read any. By 2022, those percentages had flipped.”

The recent data about the decline of reading and writing seems to support the anecdotal evidence — there is some merit behind the warnings of finger-wagging professors. Not only is the problem real, it is getting worse and will have a measurable social impact. As far as I am able to tell, A.I. will further chip away at reading and thus critical thinking. For if we believe reading makes us smarter, the use of A.I. stands to make us dumber.

In an article that was published in New York Magazine,⁴ journalist James Walsh discusses the cognitive impact of A.I. It is worth quoting at length:

Some early research shows that when students off-load cognitive duties onto chatbots, their capacity for memory, problem-solving, and creativity could suffer. Multiple studies published within the past year have linked AI usage with a deterioration in critical-thinking skills; one found the effect to be more pronounced in younger participants. In February, Microsoft and Carnegie Mellon University published a study that found a person’s confidence in generative AI correlates with reduced critical-thinking effort.

The fear is not that we will devolve into a pre-Gutenberg world without widespread literacy. The fear is we will evolve into a post-literate society wherein our ability to read and concentrate on long-form, interesting, intellectually challenging, and nuanced writing will be compromised and supplanted by other information delivery systems (online media, A.I., etc.) that don’t stimulate the brain in cognitively beneficial ways.

Will A.I. kill books, reading, and the publishing industry? No. That’s not my argument. In the short- and mid-term, reading culture will suffer, as will the publishing industry and independent web publishers. In the long term, we eventually may see a revival of reading in the same way that digital music products (mp3s and streaming) has revitalized vinyl sales. Rather, my argument is that the use of A.I. discourages reading and the reading of books, and its societal implications are cause for serious concern.

This is not to say that there are no use-cases for A.I. Transcribing meetings and interviews, entering data, aiding healthcare professionals in diagnoses, and helping us conceive of new ways to solve problems are among the use-cases that offer utility without compromising reading and cognition. Yet using A.I. in ways that shoulder our cognitive burden — to ask it to do our reading, writing, and thinking for us — doesn’t bode well for the future of our society.

If A.I. impacts readers, it will impact writers, journalists, and those who are committed to good writing and the pursuit of truth. It will impact the publishing industry, from newspapers and magazines to books. This is not about production. It is about consumption. Unless we do more to resist smartphones, algorithmically mediated feeds, and artificial intelligence, we may see a readership, and all the indicators of an intelligent society that go with it, decline.

But A.I.’s impact might be felt among certain communities more than others.

In his famous essay “What is Enlightenment?” the philosopher Immanuel Kant remarks that laziness and cowardice are among the reasons why people don’t think for themselves. But I wonder if time — or the lack thereof — might also be a reason why people don’t think for themselves. Students often report not having the time to read and complete assignments. This is especially true for the population of students at my institution, many of which have jobs, children or other dependents to care for, unreliable transportation, or external pressures that reduce the amount of time they can spend on school. Students from affluent households have more time to read and complete their assignments because they don’t have the same external pressures as those from less-wealthy households. The fear is that students from impoverished households will rely more on A.I. tools than those from affluent households, further widening the education (and socioeconomic) gap in our society.

What can we do?

If you’ve made it this far, you’re likely not part of the problem. And you’re likely someone who is concerned with literacy, books, and with reading, writing, and thinking. For those of us who are concerned with artificial intelligence, we might do well to be more vocal about the consequences of using it: intellectual atrophy. We need to discourage the use of A.I. in the processing of information. We need to encourage people to read more high-quality, nuanced, and challenging work and less of the A.I. slop that seems to be proliferating online.

Educators might consider developing solutions and incentives that promote reading and ensure students cannot use A.I. on their assignments (like in-class assessments, low-tech classroom environments, close readings, etc.). This opinion is not new. Jeff Schatten in his 2022 article mentioned above speculated that A.I. technologies would get so good that, if we wanted to ensure our students are submitting original work, we would need to revert to assignments and tasks that don’t rely on technology. But let’s not confuse the educational enterprise with production. Public education is less about producing a paper, conducting research, and churning out “evidence of learning” than it is about asking questions and developing skills like critical thinking, a task made difficult — if not impossible — by the overuse of A.I.