Listening to Skepticism: What Faculty Concerns About Generative AI Reveal

Across higher education, conversations about generative artificial intelligence (GenAI) often begin with questions about how to adopt it, where to apply it, and how quickly to do so. For many faculty, however, that framing skips over a more basic question: Should this be used in my teaching at all and, if so, under what conditions? And yet, faculty are being asked to decide whether and how GenAI belongs in their teaching at a moment that already feels destabilized. They are doing so while facing internal pressures to adapt courses and clarify classroom policy, as well as external pressures created by the rapid proliferation of tools, punctuated by bold vendor claims and contentious debate about the value of those tools. Financial pressures in higher education make efficiency-oriented promises particularly salient—as do broader expectations that colleges and universities keep pace with AI—even as fundamental questions about GenAI remain unresolved.

For some, GenAI presents real opportunity. It may help streamline routine tasks and create more room for work that feels less task driven and more purpose driven. For others, however, it introduces new challenges, including the need to evaluate yet more tools and methods, revise tried-and-tested assignments, and set new boundaries for student use, enforced through an increasingly sophisticated surveillance system—all within the limited time available to faculty.

In such an environment, the conversation can easily become dominated by an overwhelming surge of recommendations, resources, and calls for adaptation or adoption. But before institutional leaders ask faculty to do more with GenAI, a more useful first step might be to ask a simpler question: What exactly are faculty worried about?

The Value of Listening

As conversations about GenAI in teaching intensified at Augsburg University in fall 2025, I began this small listening project to better understand how faculty were making sense of the technology in relation to student learning. My goal was not to measure adoption but to hear more directly what concerns faculty held, what conditions shaped acceptable use, and what kinds of support might make careful experimentation possible. This project was built on local surveys and was informed by faculty AI survey findings reported by organizations such as EDUCAUSE and the Digital Education Council.^Footnote1 I gathered data from sixteen faculty members: eight from the School of Natural Sciences and eight from the School of Humanities and Social Sciences.

Listening matters because it brings to light the reasoning behind concerns, not just the concerns themselves. Using a common set of short questions, I conducted fifteen-minute listening sessions to explore faculty concerns about GenAI in their work, the conditions under which limited or acceptable use might feel appropriate, routine or low-risk tasks they might consider exploring and how they would judge success, as well as the supports they would need to experiment with GenAI for their teaching. These sessions were “notes only” (not recorded), and responses were anonymized.

Where Concerns Clustered

Overall, the sessions included faculty across a spectrum of AI stances, from current users (37 percent) to those who did not plan to use AI (25 percent), while the remaining faculty described themselves as either curious (19 percent) or skeptical (19 percent). Two patterns stood out. First, faculty stance toward AI did not map neatly onto work-related use. For instance, among faculty open to GenAI, 58 percent were already using AI in their work; those not currently using AI in their work expressed hesitation that was conditional rather than categorical—they were open to limited use, for example, but only once they had identified lower-stakes tasks for which experimentation felt manageable. Second, the concerns that surfaced were not scattered but clustered in a few recurring areas.

Between 88 and 94 percent of faculty expressed concerns surrounding student learning outcomes, disciplinary fit, and the risk that students might offload too much cognitive effort to GenAI. Across conversations, faculty repeatedly returned to the question of what students are actually learning when GenAI begins to take over aspects of the thinking process that higher education is meant to scaffold and develop.

In several sessions, faculty described this as a form of premature automation in which students might produce polished work without engaging in the mental processes those tasks are designed to cultivate, potentially eroding student development. For example, one faculty member worried that GenAI might redirect student efforts toward the production of easy outputs rather than engaging in the academic “struggle” (i.e., putting in the necessary intellectual work that builds a lasting base of knowledge):

“I worry that when language is reduced too quickly to pattern completion, thought itself risks being flattened. And when thought is flattened, diversity of thought is diminished.”

Relatedly, when discussing learning outcomes, faculty were especially concerned about critical thinking, which many considered the central capacity linking multiple core aims of higher education. Specifically, they worried that in using GenAI, students are simply being handed answers, summaries, or polished products without reasoning, iterating, and reflecting.

At a basic level, one faculty member witnessed students being unable to explain what they meant or the reasoning behind work they just submitted. Another framed this concern in civic terms, arguing that GenAI may hamper students’ ability to learn how to form and communicate ideas in ways that foster informed judgments, potentially leaving them susceptible to misinformation, disinformation, and conspiratorial thinking. Some went further still, raising concerns about metacognition, suggesting that GenAI may weaken students’ ability not only to think but also to monitor and direct that thinking in ways that build on, extend, and synthesize their prior learning:

“The real risk isn’t that AI makes student work easier. It’s that AI lets students bypass the hidden layer of learning—the place where thinking is formed, not just expressed.”

Faculty also questioned whether GenAI is appropriate across the spectrum of educational contexts, from broad disciplinary aims to specific tasks, assignments, and pedagogical values. For some, this was not just a practical concern but a deeply personal one because it seemed to sever the educational compact between instructor and student and displace the very work that teaching is meant to support.

In the humanities, a faculty member questioned whether GenAI belongs in the classroom, where voice, uniqueness of interpretation, and human expression are central. A natural sciences colleague was more open to GenAI use for enhancing workflow but drew the line at any use that would replace the cognitive effort necessary for creative inquiry, writing, and the interpretation of scientific results. To many faculty members, these classroom tasks and activities were not merely outputs that checked a curricular box. Instead, they were discussed as a means for ideation, comprehension, and expression:

“I feel that AI robs me of my contribution to teaching—my place in the teaching process—as well as learning.”

A distinct but related set of concerns involved matters of ethics (e.g., integrity, plagiarism), accuracy and reliability, and transparency/disclosure. These issues were especially prominent in humanities-oriented interviews. For instance, one faculty member called for a reimagining of plagiarism and transparency policies due to GenAI and wondered, if AI is used, who is actually producing the work, on what grounds that work should count as legitimate, and whether GenAI use should be disclosed explicitly whenever it shaped a submission.

Another faculty member described GenAI use as potentially breaching an “unspoken contract” between instructor and student, especially when assignments are carefully designed to support student development and rely on good-faith efforts. For others, those concerns touched a deeper question about educational authenticity—how to preserve genuine human input in teaching and learning:

“I do not want a generic, impersonal voice entering spaces where students should be developing their own voices and thoughts.”

Institutional Supports Faculty Requested

Most faculty interviewed did not ask for sweeping institutional changes before engaging with GenAI. Instead, they pointed to a more modest yet concrete set of supports such as clearer parameters and policy guidance, time, and more formal institutional incentives that allow for careful and responsible experimentation. They also called for multilevel supports (including IT help) to allow GenAI use to feel structured, understandable, and manageable, not spontaneous or makeshift.

For some faculty, the key issue was not whether GenAI should be permitted in the abstract but whether they could rely on a framework for permitting, enforcing, and evaluating its use in ways that were realistic, student-centered, and pedagogically sound. Others emphasized that the real barrier was capacity in practice—learning GenAI requires time, design energy, and support that many faculty do not currently have. Some faculty also raised the need for institutional infrastructure itself, including strong governance, privacy protections, and access to well-supported tools:

“The issue is not only having clearer rules but whether those rules can be implemented in a realistic, humane, and sustainable way.”

“Careful scaffolding [with AI-incorporated assignments] requires faculty time and energy that many of us do not have.”

Peer and Departmental Supports Faculty Requested

From their peers and departments, faculty frequently asked for structures that would make experimentation feel legitimate, valued, and manageable. For instance, they sought a cohort model rather than testing or improvising with GenAI alone. They called for departmental conversations not only to discuss whether GenAI is appropriate but also to consider specific use cases in their labs, language learning, writing-intensive assignments, or other contexts for which being fully present in the learning process was central. Others wanted to learn from colleagues already experimenting with GenAI so that they could anticipate the likely “landmines” before introducing the technology deliberately into their own classes.

Faculty also expressed interest in small, targeted pilots described as tightly framed exercises in which GenAI could be the actual object of study or critique rather than a substitute for instruction. Examples included assignments for which students tested GenAI output for accuracy, discussions about the ethics of this new technology, compare/contrast activities using AI-generated versus human writing, or small, class-specific modules in which GenAI predictions were evaluated against actual results. These were acceptable because they preserved the instructor’s control over the learning process and did not displace the core cognitive work students were supposed to do. (See the sidebar “Three Small Supports.”) That said, faculty pointed to common limiting factors such as the lack of time, front-end design and scaffolding, and the challenge of monitoring AI-related work effectively and responsibly. Two responses captured this tension well:

“What would help is collective support, not just individual permission.”

“I do not have time to front-load an experiment [with AI] that may lead to fruitless rabbit holes.”

Conclusion

The overall patterns that emerged from listening to faculty perspectives about GenAI use suggest that their concerns and skepticism were not kneejerk rejections of any change. Instead, they were asking to be heard and supported before being asked to adopt or adapt, a reasonable request in a moment that to many feels unreasonable.

These brief listening sessions surfaced not a scattered set of reactive grievances but a reasoned, thoughtful collection of concerns from faculty who deeply care about their students and the stakes of potentially getting it wrong with GenAI. Many are willing to engage carefully with the technology if the conditions reflect the gravity of their concerns, which clustered around student learning and development, core educational values, and the integrity of the teaching relationship. These concerns came paired with concrete, actionable requests that faculty were asking of their institution and departmental colleagues.

Several themes that emerged here align with the patterns reported in broader higher education surveys. Recent EDUCAUSE research similarly highlights faculty concerns about GenAI impacts on independent thought, absence of clear best practices, and policy uncertainty.^Footnote2 What this project adds is texture—a closer view of how faculty think through their concerns and what they believe would help them. With faculty making up just 12 percent of EDUCAUSE survey respondents, complementary approaches that center faculty perspectives directly are worth pursuing.

Given the rapid shifts in the AI landscape, the window for getting this right may be narrowing. Institutions that move quickly to adopt without first understanding what faculty are navigating risk deepening the uncertainty they are trying to resolve. Listening is not a delay but a precondition for responses that are precise, trustworthy, and respectful of faculty time. That, ultimately, is what these faculty were asking for.

Notes

1. School of Natural Sciences Faculty AI Survey, internal report, Augsburg University, July 2025 (n = 45), unpublished, anonymized summary on file with the author; Center for Teaching & Learning Workshop Live Poll, Augsburg University, August 28, 2025 (n = 22), anonymized session results on file with the author; Jenay Robert and Mark McCormack, 2025 EDUCAUSE AI Landscape Study: Into the Digital AI Divide, research report (Boulder, CO: EDUCAUSE, February 2025); and Digital Education Council, “Global AI Faculty Survey 2025.”Jump back to footnote 1 in the text.↩
2. Jenay Robert, The Impact of AI on Work in Higher Education, research report (Boulder, CO: EDUCAUSE, January 2026). Jump back to footnote 2 in the text.↩

---

Henry Yoon is Associate Professor and Program Lead for the AI-Informed Assignment Pilot at Augsburg University.