The Faculty Retreat From AI Has a Student Cost

Delaware County Community College’s Susan Ray, PhD, offers commentary on how the faculty retreat from AI has a real student cost based on two new surveys. This article originally appeared in Insight Jam, an enterprise IT community that enables human conversation on AI.

The Retreat

“I just remembered one of the first days of class,” one of my first-year writers shared in her final course reflection, “when I heard that we were going to be using AI. I was a bit shocked knowing that most teachers absolutely hate AI and are so opposed to students using it.”

Her initial resistance faded as, over fourteen weeks, we moved from fill-in-the-blank prompting to students conducting independent research with LLMs, giving them repeated opportunities to test and evaluate these tools for themselves. She had entered ENG 100 unsure about AI after “a few bad experiences” with it over the years. By the end, she recognized that it could help in moments when she was “stuck or need[ed] help trying to get my foot on the gas.”

She was by no means a converted AI enthusiast by the end of our course, but instead someone far more valuable in this complicated moment: a user who approaches it with skepticism and discernment. “It’s all about how we use it,” she concluded.

This spring I taught four sections of ENG 100 at Delaware County Community College, where I’ve been an English professor for thirteen years. For the past two, I have integrated AI literacy into my writing curriculum, teaching students to use large language models critically, ethically, and transparently. They maintain AI Transparency Journals, documenting every course-related exchange, evaluating outputs for hallucinations and bias, and verifying claims against outside sources. In class, we hold sustained conversations about how AI is reshaping the workforce they seek to enter, measuring it against the technology’s very real environmental and human costs.

First-year college students arrive on our campuses with AI already built into their phones, their feeds, the tools they use to search and write. They see its effects, but familiarity does not equal fluency. Thus, we owe them spaces where they can ask how these systems work, what they cost to run, and whose labor and language they draw from. And space to practice the harder skill: keeping their own thinking at the center even when help is a click away.

Yet each semester, more faculty are refusing to engage with these technologies, leaving students to figure them out on their own, vulnerable to companies that market their tools as magic, or to peers who may know little more than they do.

The Digital Education Council’s AI in Higher Education Global Survey 2026 gathered 45,398 responses from students and faculty across 35 countries, and one finding in particular should shock North American professors. While intent to use AI in their teaching remained relatively stable among faculty in Asia, Europe, and Latin America, in the United States and Canada it dropped nine percentage points, from 76 percent to 67 percent, within a single year.

And ours is the lowest projected adoption of any region surveyed.

A second survey, published in January by the American Association of Colleges and Universities and Elon University’s Imagining the Digital Future Center, uncovers the motives behind this mass retreat. Of the 1,057 U.S. faculty surveyed, 90 percent said generative AI will erode students’ critical thinking. Nearly half think its effect on their students’ careers will be more harmful than beneficial, and a quarter admit to never having used a generative AI tool. This number rises to a startling 40 percent among arts and humanities faculty.

Two surveys. Opposite sides of the lectern. One visible retreat.

My argument here is a modest one: no student should graduate without having spent real classroom time learning how these systems work and how to use them with judgment. And every faculty member should understand how AI is reshaping the practices and assumptions of their own discipline. These surveys show us what happens when institutions do neither.

The Vacuum We Made

Let’s start with the trust deficit. Only 17 percent of students in the United States and Canada believe their instructors are prepared to guide them in using AI for learning, and the AAC&U data helps explain why: 68 percent of U.S. faculty report that their institutions have not prepared them to teach or mentor students with these tools.

A starker measure of this widespread deficit is tucked between the lines of the AAC&U report: while 87 percent of faculty have written their own AI policies for students, only 48 percent say their institution has developed campus-wide guidance, and just 35 percent report that their department has created policies tailored to their discipline.

When we step back and take stock of this gap, we find nearly nine in ten faculty are setting their classroom policies alone, many without the baseline literacy to ground their decisions, leaving them to improvise the governance their institutions never provided.

Understanding how AI is already molding our fields beyond the classroom, and using that knowledge to decide where it belongs in our curricula, requires time and sustained discussion. Because the professions our students are entering are demanding thinkers with enough disciplinary expertise to recognize when AI sharpens their work and when handing over the thinking means surrendering the very expertise their education was meant to develop.

Much like our students, we cannot become fluent in a new technology through a handful of brief encounters. Fluency develops through application and scholarly inquiry. To accomplish this, faculty need guidance, resources, and, perhaps most importantly, time, the most precious resource of all. We need more hours to experiment, to reflect, and to learn from one another.

And once faculty better understand these tools, the question is no longer whether to allow AI into their courses, but whether they can defend why not. Two teachers can ban AI on an assignment and mean two opposite things by it: one has weighed what the tool would do to the thinking the assignment was built to produce while the other never opened the browser window to evaluate the merit of their decision. The policies read the same, but only one of them is teaching with intention. Refusal can be the most rigorous choice in the room, but ignorance can arrive dressed in the same clothes.

What it Costs Students

We owe our students the benefit of our informed judgment, but instead we’re graduating them with the inheritance of our fear. The student numbers capture it, and they are more disturbing than any measure of faculty resistance. The DEC tells us forty-three percent of students in the United States and Canada say they would support a campus-wide ban on AI in academic work, while seventy-three percent worry that classmates may be using it themselves as an unfair advantage.

We should not read these numbers as vindication. They show what happens when students inherit our uncertainty without a framework for judging the technology themselves. Since ChatGPT entered the classroom, many have heard AI discussed primarily as cheating, danger, replacement, and decline. Their suspicion is not paranoia; it is the predictable result of classrooms without shared norms and a professoriate still unable to agree on where use becomes misconduct.

When we return to the AAC&U survey, we see that 52 percent of faculty say a student who follows a detailed AI-generated outline is cheating. Yet, 45 percent say it is okay for a student to write a paper, run it through an AI system, and make the suggested changes. In nearly every scenario, large numbers answered “not sure”—this mass, uninformed uncertainty that confuses our students and leaves them ill-prepared for the AI-infused world outside our classrooms.

When the rules are unwritten, students draw their own boundaries in private, then glance at each other across the classroom. We created this vacuum of guidance, and it has filled with mistrust.

Meanwhile, we’re graduating them into AI-enabled medicine, AI-assisted accounting, AI-driven legal research, and AI-supported writing, teaching, engineering, and design. Handing them our anxieties is insufficient preparation.

The Digital Education Council survey uncovers a troubling insulation. Fifty-eight percent of faculty in the United States and Canada say they are not worried that what they teach will be outdated by the time their students graduate—the highest confidence reported in any region. But just 19 percent of students in those countries believe their programs are current enough to prepare them for an AI-shaped future—the lowest student confidence in the world.

The AAC&U findings take away the last comfort, because there the faculty indict the system themselves. Sixty-three percent say last spring’s graduates were not ready to use generative AI on the job, and 71 percent say those graduates did not understand the ethical questions these systems raise.

We know the students crossing our commencement stages are unprepared. We also know whose job it was to prepare them.

Awareness, Not Adoption

So what am I asking of the professor who decides not to teach students how to use AI?

Not integration. Awareness.

The institutional failure is just as quantifiable. Only 13 percent of faculty in the AAC&U study report that their institution has adopted AI literacy as a general-education learning outcome, and 70 percent say they lack the training and support infrastructure to bring these tools into their classes.

Much as I would like to hand this work to individual faculty to sort out in their own classrooms on their own time, it is not really theirs to carry. Preparing students for a world increasingly shaped by AI is an institutional job. It requires sustained, tiered faculty development, and an institutional commitment to AI literacy as real and solid as the commitment to writing, or statistics, or calculus: embedded in program outcomes and general education, examined through assessment, and taken as seriously as any other learning goal.

For the institutions currently considering campus AI platforms, the order of operations matters: literacy investment has to precede tool procurement, or the tools land in exactly the vacuum these surveys describe. The real evaluation starts before the demonstration: What learning outcome will the tool support? What student data will it collect, retain, or reuse? How will the institution address accessibility, privacy, equity, and faculty preparation before deployment? Answering them well means resisting the easy binary of good or bad in favor of the harder, intellectually rigorous work of holding several truths at once.

The novelist Elif Shafak, speaking on the role of the intellectual, urges us to “resist this tendency to simplify things” and to “abolish that narrative that tells us you can only be one thing.” We cannot let anxiety flatten us into singular ways of seeing the world or ourselves. This multidimensional thinking is the very work of professors. The ban-or-embrace battle over AI is exactly the kind of false binary she has in mind.

Our students deserve teachers who know the world they are entering, just as we know the world that formed us.

We can’t build that capacity in a semester. But it has to be built, and the clock started three years ago.

Another student ended the same semester unmoved. “You’ve not changed my stance on being pretty heavily anti-AI,” they wrote, “and I stand by it . . . but . . . Nuance. Hard conversation. It’s kinda here and we gotta deal with it.”

That student stood by their objections. They stayed in the room. They engaged anyway.

Fresh out of high school, many of them have already done what too many of us with graduate degrees and decades of scholarship have failed to do. Their future hinges on our choice—not to embrace AI, but to stop retreating from the responsibility to understand it.