In a recent national survey, we asked college instructors to share the AI-related skills they prioritize and teach in their classes and compared their responses to employers’, using an AI skills Framework originally developed by researchers at the multi-national human resources technology company HiBob. Asking instructors to prioritize AI-related skills within a specific framework inevitably begs the question of what other skills those instructors consider important. When offered the opportunity to suggest additional skills, many instructors noted how thorough the existing framework was, with one respondent saying that framework “was quite comprehensive in its scope.”

However, there was also an emphatic collection of responses that suggested the most important AI-related skill was to not use generative AI at all. These sentiments went beyond the skill of knowing when human or AI capabilities are best deployed or, as one instructor noted, “that AI shouldn’t be used for everything.” A small, but vocal, contingent of instructors framed the avoidance and rejection of AI in its entirety as a skill or mindset to be cultivated in college. Although they agreed on not using AI, these opinions are not monolithic. The nuances the survey uncovered may help academic leaders better understand—and potentially address—faculty perceptions of AI.

We found that instructor AI resistance fell into at least one of the following categories, with many responses reflecting aspects of several categories:

  • AI interferes with learning. These responses did not object to AI in and of itself, but focused on the negative impact student use of AI has on learning. Instructors noted that students “should be able to do things without any assistance from AI such as analyze data/reading on their own” and emphasized the value of “having the student learn through doing.” While an important thread within the responses, worry about the impact of AI on learning is not a resistance to AI in principle. Instructors expressing these sentiments worry about when AI usage crowds out important content and skill acquisition as distinct from AI use being inherently wrong. One respondent called on students to understand “when the use of an AI is removing the learning of skill” rather than issuing a blanket admonition not to use AI at all.
  • AI undermines cognitive development. The concern about cognitive and intellectual development is more expansive than that about its impact on learning. Instructors expressing this kind of resistance worry about more foundational capacities like thinking, attention, and perseverance through difficulty. One instructor quipped that the necessary skill for students is “How not to let AI replace their thinking,” and another suggested that students should be “able to think for themselves without becoming dependent on AI tools.” This aspect of resistance cuts close to the core of academic identity, or what many faculty believe about their role as educators. As one respondent argued, students should learn “habits of sustained attention, critical thinking, and real connection to other people in the absence of any use of AI or intermediary device[s].”
  • AI conflicts with disciplinary values. Faculty embrace or rejection of AI has a particularly disciplinary flavor, a pattern that emerged in our data as well. Many instructors cited their disciplinary (i.e.: art history) or domain affiliation as the source of their objection to AI. It is beyond the scope of our data to draw broad conclusions about how different disciplines view AI in teaching, but the observation that all of the AI-skeptical responses that included disciplinary information were from arts and humanities instructors echoes other research on AI resistance. One respondent summed up these sentiments succinctly, writing, “I teach the humanities. AI has no place in this work.”
  • AI is inherently unethical. Instructors cited a variety of ethical objections to AI, including environmental concerns (“Graduates should know that using AI is harmful to the environment”) and financial concerns (“It is essential students understand the ways AI takes content from authors and other creators without their consent and compensation”). The data collected for this project are insufficient to fully explore the myriad ethical objections instructors may have to AI, but it is important to note that ethical objections may be discrete or more global in nature. As one instructor observed, “There will be no large-scale incorporation of AI that doesn’t steal from creators, warm the earth, and contribute to societal brain drain.”
  • AI output is untrustworthy. This criticism of AI in teaching is less ideological than it is practical because, as one respondent overserved, “It is also critically important for students to fact check every single output from an LLM, b/c they are so totally prone to fabricating information.” The continued problem of AI reliability interacts with students’ status as novices in their academic disciplines in ways that heighten instructors’ concerns. As one instructor lamented, “These students don’t know anything.  They are using AI blindly. They need to develop non-AI expertise so that they can effectively use AI.” These instructors express a view closer to “not yet” than “never” about student use of AI.

Our analysis highlights that instructor resistance to AI in the classroom, and in higher education in general, is complex and multifaceted. Institutional initiatives around AI need to address the specific objections raised by instructors rather than dismissing them or reducing the debate to a false binary. More research is required to understand the prevalence and strength of these, or other, aspects of AI resistance; such work is essential to understanding the varied reasons why instructors may object to, or support, using AI in their work.

Such work could begin with a mixed-methods approach to validate the different dimensions of resistance through more rigorous qualitative work, followed by a broader survey exploring how strongly instructors feel about each dimension. Doing so would allow for a more granular approach to engaging faculty about AI.