By Kevin Yee
University of Central Florida

Higher Education has been in a period of unprecedented disruption, essentially nonstop, since early 2020. That is a long time indeed to face what educational futurist Bryan Alexander (2026) might describe as an early version of a polycrisis. The rapid and complete campus shutdown in order to “flatten the curve” of Covid infections was followed by a headlong rush into emergency online teaching, then a frantic conversion to “quality” online courses for everyone, and finally a return to campus. But that campus return was anything but business as usual. In many campuses in Florida, faculty were forced to make yet another adjustment and teach in a HyFlex modality: addressing a sparse, socially-distanced in-person audience at the same time as a larger synchronous online audience. Students, accustomed to more than a year of faculty extending grace (as their administrators urged them to do), continued to seek favors and customizations, contributing to an ever-growing mental health crisis. I can think of no worse moment in history for large language models (LLMs) to go viral, but they did just that in late 2022. Suddenly, “disruption” seemed too gentle a word to describe the impact on student learning and faculty frustration.

Almost all students in 2023 could be classified as AI “innovators” in Rogers’ Diffusion of Innovations (2003)—ahead even of the “early adopters” category. The reasons are not difficult to discern: LLMs offered a rapid, free, largely undetectable means of completing homework and take-home assignments. Most students found the opportunity to have more time to play, sleep, or even work their part-time jobs too tempting to pass up (and others, we have to admit, because they were overwhelmed, multilingual, anxious, or academically underprepared). At the same time, the majority of faculty on our college campuses could be considered “laggards” in that first year. Many were completely ignorant of the existence or threat to education that LLMs represented. Most faculty seemed to want to continue teaching like it was 2019: pre-Covid and pre-LLMs. This is understandable, because the alternative is the very heavy lift of redesigning assessments and the use of in-class time, as well as everything in between. When it came to LLMs in particular, they looked for ways to detect or prevent AI use, neither of which is particularly viable as a strategy.

In the meantime, today’s college students have experienced 3+ years of using LLMs with impunity. Most have not been accused of improperly using AI, let alone been formally disciplined for it, so it is not surprising that national surveys and studies like EAB (2026) continue to show accelerating student use. Faculty, even those who were laggards in 2023, have been noticing the increasing student usage and are in turn becoming increasingly frustrated with what they perceive as “AI slop” in the papers and homework turned in. Interestingly, as the years tick by, more undergraduate students are also joining in the backlash—some of them for reasons of wanting to carve out the primacy of human creativity, others in protest for the assumed negative future impact on jobs, and still others who recognize that they are attending college to train their brains, and these shortcuts, however useful at the moment, will not serve them long-term in their careers. The vast majority of faculty share this mistrust of cognitive offloading and its impact on critical thinking (Watson, 2026). The comforting reality is that some students use AI conscientiously in ways to boost learning, not replace it.

It would be one thing if growing impatience with AI were the only side of the story. But faculty (and student) perspectives differ very much on AI’s future role in society, in learning and training, and especially in the workforce. Certainly mass media advances the narrative that AI is starting to replace jobs (Tyrangiel, 2026), or at a minimum, that tech CEOs are engaging in “AI-washing,” or blaming regular layoffs on the convenient scapegoat of the day (Rogelberg, 2026). It’s hard to completely avoid the pro-AI enthusiasm of many CEOs, however, most of whom reiterate in national surveys that they now prioritize AI fluency in recent college grads over almost any other quality (Microsoft & Linkedin, 2024). Then we have the irony that while most faculty oppose student use of AI in their work, some will use AI in their own faculty work (making quizzes, lesson plans, etc.), implying that some faculty don’t see a problem with AI as part of the workflow.

This juxtaposition puts college instructors in an awkward position. Surely, if our students’ future employers expect AI fluency, we can’t just leave that to chance, or expect that students will simply pick up what they need along the way. The principles of backward design (Wiggins-McTighe, 1998) dictate that we plan and execute with intentionality toward the desired outcomes. Yet “leaning in” to AI fluency as a discrete course objective seems to contradict the goals of minimizing cognitive offloading. That goes double for allowing, or even encouraging, students to use AI as a full creation partner. Yet if we never do that, are we setting up students for failure when their first exposure to full co-creation happens on the job?

Many faculty are increasingly coming to grips with the idea that AI is here to stay and therefore can’t be just denied and banned, but their realization still leaves a mighty disconnect about what to do with diametrically-opposed mandates. Our webinar on September 8 will attempt to navigate the Scylla of academic rigor and the Charybdis of employer needs. We face significant headwinds. We can’t just ignore AI, but neither does it seem academically safe to “lean in” completely, and allow students to use LLMs to complete all the work presented to them in college. What we need is a system that incorporates “all of the above” (multiple-choice assessment pun fully intended). The answer is not universal AI adoption, nor universal prohibition. Our emerging framework intentionally creates different learning spaces: some activities where AI is encouraged, some where AI is constrained, and some where independent human performance remains essential. The goal is to produce graduates who are both capable thinkers and capable AI users. At the webinar, we’ll discuss the system we evolved at UCF that allows for both—hopefully you can join us on September 8!

Register for upcoming FALCON events

AI use disclosure: After this article was human-drafted, it was pasted to ChatGPT 5.6 (July, 2026) to identify problematic phrases or grammar, and to verify that the declared main points were adequately explored.

Works Cited

Alexander, B. (2026). Peak higher ed: How to survive the looming academic crisis. Johns Hopkins University Press.

EAB. (2026, February 24). How students view and use AI in college search: Insights from a survey of 5,000+ high school students. https://eab.com/resources/insight-paper/how-students-view-and-use-ai-in-college-search/

Microsoft & LinkedIn. (2024, May 8). AI at work is here. Now comes the hard part: 2024 Work Trend Index annual report. https://www.microsoft.com/en-us/worklab/work-trend-index/ai-at-work-is-here-now-comes-the-hard-part