The End of the Essay? Humanities Departments Enter the AI Age

We call it hell week.

At the close of each semester at my small liberal arts college, students bury themselves in the library, staking out study rooms, flipping through stacks of books and notes, and squinting at laptop screens. Term papers are due at the end of the week, and extensions are hard to come by.

The academic writing process, which makes up the bulk of undergraduate education in the liberal arts and humanities, now stands in the crosshairs of developing artificial intelligence as widely available large language models (LLMs) continue to improve at replicating human thought and spitting out strings of text aggregated from internet resources.

The technology stands to greatly improve many facets of research and industry but threatens to leave one educational institution — the undergraduate college essay — in the dust.

Students in high school and college are using chatbots to write essays, whether their teachers like it or not. New research from the Pew Research Center shows the share of teens who said they use ChatGPT for their schoolwork doubled in a year, rising from 13 percent in 2023 to 26 percent in 2024. Eighteen percent of those surveyed said they believed it was academically honest to use the software to write essays. But because students have been passing off the work of others as their own since the Roman poet Martial first called his rival Fidentinus a “plagiarus,” it stands to reason that more than 18 percent will do it anyway.

Even the more conscientious students may soon struggle to write for themselves as companies embed LLM technology into email platforms, word processors, and writing-assistant software. As I type, Google Gemini occasionally suggests the possible next phrase as a phantom in front of my cursor.

In the AI age, college humanities departments face a choice: Either openly incorporate AI into class writing assignments or ban the technology and figure out a way to prevent students from using it.

To find out more about how colleges are handling the problem, I talked to a handful of humanities departments and asked how they are — or are not — incorporating the use of AI into humanities courses and academic writing assignments. The professors I spoke with spanned the spectrum of tech adoption, from instituting no-tolerance policies to actively working to incorporate AI in the classroom. They all agreed, however, on what’s at stake: At its most fundamental level, writing is thinking — something chatbots can only simulate but threaten to displace.

For Suzy Paalman, dean and interim president of St. John’s College in Annapolis, Md., the use of AI in the writing process is out of the question.

“We see our classroom writing as a human activity, meant to help us develop our own questions and thinking on topics of fundamental importance,” Paalman said.

Including passages composed, revised, or translated by AI in one’s written work falls under plagiarism in the college’s policy on academic honesty. Even using a chatbot to brainstorm topics for a paper circumvents the important intellectual work at the center of the writing process, according to Paalman’s assessment.

“Technological aids need to be carefully considered . . . that they not impede the learning process, including the process of figuring out what question the writer wants to ask,” Paalman said. “The more developed the ‘help’ is, the more likely it is to circumvent the, at times painful, process of one’s own inquiry. In our judgment, generative AI may appear to be helpful but, in fact, is not.”

The college faculty produced a report in the spring of 2024 on how the college views essay-writing and the role that large language models would play in that process. The essay is the product of a continual process of intellectual inquiry, the report said. The process is worthwhile not only because it yields a product but because it helps the writer to achieve self-knowledge.

“The seemingly omnipresent offer of artificially generated assistance is false, intrusive, and dangerous,” the report said. “It promises to supplement but actually supplants your writing. It deprives you of the rare and beautiful chance to think. It makes you show a false face to the world and does you no real good. It produces something no one wants to read.”

In short, writing cannot be outsourced because thinking cannot be outsourced.

Whether professors want to regulate or incorporate the use of AI in their classrooms, learning to use the technology is the first step, according to Leo Lo of the University of New Mexico. Lo, one of the scholars at the forefront of humanities research into AI, describes himself as an advocate for AI literacy.

“My approach is — whether you’re for AI or against AI, you can make your argument stronger if you know more about AI,” Lo told National Review.

As dean of the college of university libraries and learning services, he started a summer program at the public university for humanities professors and graduate students who want to learn more about the capabilities of the technology and how to incorporate it into their teaching. The program gives each participant access to ChatGPT Plus and allows them to experiment with the technology over the course of eight weeks within a cohort of fellow academics from universities across the country. The structure allows instructors to play with AI during the summer and become comfortable with the technology before introducing it in their classrooms.

Artificial intelligence may be incorporated effectively if professors are clear about standards for academic integrity, Lo said. While students should always be responsible for the accuracy of their work, he said, AI can help students to brainstorm, generate ideas, and draft outlines.

Most likely, all writing in the future will be a combination of both AI and human output, he said. Rather than fixating on the technology as a threat, professors should emphasize finding ways to ensure that students develop the critical-thinking and communication skills historically cultivated through the writing process.

A middle way, suggested by Miriam Wallace of the University of Illinois-Springfield, presents LLM technology as a research assistant and advanced editing tool rather than a first-draft tool. Working under a grant from the National Endowment for the Humanities, Wallace and Emily Todd of Eastern Connecticut State University recently finished a research initiative to help humanities professors at public liberal arts universities develop curricula for students to work with AI in a humanities framework. Twenty professors from five universities across the country met once a month for a year, experimenting with the potential and limitations of generative AI and sharing their findings.

The verdict: AI is a bad writer — at least on the first draft. Everything written by an AI sounds the same and resembles the work of a first-semester college sophomore, Wallace said. The technology is much better at scanning a lot of material quickly and suggesting revisions for written work. These functions open the potential for using it as an aid for other kinds of work, such as searching historical archives, editing, or teaching second languages.

Like Paalman and Lo, Wallace suggested that the purpose of a writing assignment is not the product but the process of discovering what you think.

“If you ask the AI to do it, you’re not getting that benefit,” she said.

Professors must design assignments so students either cannot or do not want to use AI in place of thinking, Wallace said. Because “check-box” assignments such as discussion posts and five-paragraph essays invite turning to AI tools, Wallace suggested crafting assignments that involve more self-reflection or multiple steps to develop a thought.

Paalman has a slightly more traditional approach. St. John’s College already has a tradition of oral exams and “paper conferences,” discussions between student and tutor about the paper after it has been turned in. In addition to these assignments, Paalman said, some faculty have begun to do writing exercises in class in order to completely remove the possibility of LLM usage.

In-class written assignments like these will likely replace the conventional term paper in many cases as professors seek practical ways to ensure that students write for themselves. For assessing comprehension, quote identification tests, oral exams, and recitations are becoming more common. These adjustments are necessary and good.

At a higher level, however, students must be convinced of the importance of thinking.

Aaron MacLean, a former professor at the U.S. Naval Academy, took this approach in a recent speech at the Hudson Institute in Washington, D.C. Because generative AI threatens our ability to reason and organize our thoughts, it also threatens our freedom, he said. Philosophy and strategy are the means by which we preserve our freedom.

“Philosophy and strategy are not things that you know,” he said. “They are things that you do. If we cannot reason and if we have no exposure to wise teachers — often in the form of their writings — we cannot strategize or philosophize — and we will no longer be able to maintain our freedom.”

No one serious about liberal arts education, it seems, is convinced that academic writing no longer has merit in the AI age. Rather, it may be more important than ever.

MacLean put it most simply: “Having AI do your work for you is like sending a robot to the gym to lift weights on your behalf. Your brain and your muscles will experience similar results.”

As professors grapple with redesigning curricula, the real responsibility for the preservation of academic writing falls to the students. It is their ability to think that is at stake. Those convinced of the necessity of writing and willing to do the work will continue to reap the benefits of rigorous thought — even if they go through hell in the process.