What faculty actually worry about when they worry about AI
National surveys say the worry is not mainly about catching students. It is overreliance, thinning critical thinking, and workload growing faster than the guidance meant to manage it.
The Provenance Team
Faculty concerns about generative AI are sometimes described as resistance to change, but that framing is too easy and not especially accurate. Many faculty are already using AI in their own work, testing it in courses, and revising assignments to help students see when a tool supports learning and when it replaces the very work the course was designed to teach. The concern is not that technology exists. The concern is that the connection between student effort, student understanding, submitted work, and academic credit has become less visible.
That concern is showing up clearly in survey data. Elon University and AAC&U reported in 2026 that 95% of faculty in their national survey believed generative AI would increase student overreliance on AI tools, 90% believed it would diminish students’ critical thinking skills, 83% believed it would decrease attention spans, and 74% believed it would negatively affect the integrity and value of academic degrees (Elon University, 2026). College Board reported similar concerns, finding that at least 84% of faculty agreed that AI tools make students more dependent on technology and less likely to develop critical thinking skills, express original ideas, or engage deeply with course material (College Board, 2026).
What the worry is really about
Those are not small worries, and they should not be dismissed as a failure of imagination. Faculty are not merely protecting old assignments or older forms of student work. They are asking whether students are still doing the cognitive labor that courses are meant to require: reading closely, drafting, revising, debugging, interpreting evidence, applying concepts, and learning from difficulty.
That last phrase matters because much of learning is not efficient. Students develop judgment through slow, uneven work: they misread a source and learn to read more carefully, write a weak draft and discover what revision means, produce code that fails and learn to trace an error. If AI absorbs too much of that struggle too early, the submitted artifact may improve while the student’s judgment remains underdeveloped.
Faculty see this risk in ordinary course moments. A student submits fluent prose but cannot discuss the source base. A student produces working code but cannot explain a design choice. None of these proves anything by itself, and faculty know better than to treat a hunch as evidence, but these moments reveal why AI has unsettled assessment practice.
Polish isn’t the issue; it’s how easily it can conceal what’s missing.
Polish can hide what’s missing
In many fields, polished work has become easier to produce, and because the surface quality of student artifacts can rise without a matching rise in understanding, faculty must spend more time asking what the artifact actually represents. Tyton Partners reported in 2025 that, as instructors worked to make courses more resistant to inappropriate AI use and to verify outputs, 41% cited an increase in workload, up from 34% the year before (Tyton Partners, 2025). D2L’s summary of the same research also noted that instructors and administrators reported added work from monitoring student AI use, even as daily AI users among faculty sometimes reported workload decreases when using AI for course-development tasks (D2L, 2025).
This mixed pattern fits what many faculty experience. AI can help an instructor draft examples, generate practice questions, or adjust materials, but the same technology can make assessment more fragile if students use it to bypass learning. Faculty may gain time in preparation and lose it in verification, documentation, and student conversations. The pressure shifts from creating course materials to determining whether the submitted work supports the grade.
That pressure is heavier because faculty confidence and institutional guidance have not kept pace with student use. Ithaka S+R’s 2024 national survey found that 72% of instructors had experimented with generative AI as an instructional tool, while only 18% agreed or strongly agreed that they understood teaching applications of generative AI, and only 14% felt confident in their ability to use it in instruction (Ruediger et al., 2024). Those numbers describe a faculty body that is not standing still, but is also not yet well supported.
The unsupported middle is where frustration grows. Faculty are expected to set AI rules, design assignments, anticipate misuse, teach responsible use, preserve standards, and prepare students for a workplace where AI will be common — all while maintaining fairness across students whose access, preparation, language background, and AI literacy vary widely. A short policy paragraph cannot carry all of that.
Bigger than academic integrity
Faculty worry about academic integrity, but that phrase narrows the issue too much. They also worry about student development — whether students are losing patience for sustained reading, writing, and problem solving, and whether graduates can produce acceptable artifacts but not explain them. They worry about workplace readiness when students become tool-dependent without becoming more capable, and about the credibility of a credential that depends on grades still reflecting demonstrated learning.
These concerns are not anti-student. Often they come from taking students seriously enough to insist that convenience should not replace formation. Faculty know students are under pressure, and that AI can be genuinely useful: a student who uses it to generate practice questions, clarify a concept, or test a draft may learn more effectively, while one who uses it to complete the assignment without wrestling with the material receives a better-looking product and a worse education.
Assessment design has to distinguish those cases as much as possible.
The distinction cannot rest on whether a faculty member can detect AI involvement after submission. Detection tools may contribute to a review process, but they do not show what a student understands or repair an assessment that no longer provides enough evidence. Faculty need ways to connect the submitted artifact to the student’s own explanation and reasoning.
Back to the artifact
This is why the artifact should become the review surface. If a student submits a paper, they should be able to speak to a specific claim, source, or counterargument. If they submit code, they should be able to trace a function, explain an error case, or justify a design choice. If AI use was permitted, they should be able to explain what the tool contributed, what changed, and how the final artifact was checked.
These questions do not assume wrongdoing. They assume responsibility.
They also return the conversation to learning. The goal is not to make students anxious or faculty suspicious by default; it is to produce better evidence about whether the artifact represents demonstrated understanding. Students who know their work deserve a meaningful chance to show it, and faculty making academic judgments deserve a record stronger than impressions, detector scores, or after-the-fact reconstruction.
Where Provenance fits
Provenance Learning is designed to take those worries seriously without treating them as panic. It does not ask faculty to accept a polished artifact on trust, and it does not treat every student as a suspect. Instead it gives faculty a structured way to ask the questions they already know are educationally meaningful — can the student explain the work, connect choices to evidence, account for any tool use that was allowed? Faculty approve the questions and criteria and keep the final judgment; students answer prompts grounded in the assignment, outcomes, instructions, and specific features of the work; and the session records audio, timecoded transcription, artifact references, and observational notes, plus evidence-linked recommendations faculty can override. The concern underneath all of it is not the tool. It is whether the student is still developing the understanding the credential is supposed to represent.
Faculty worry about AI because they know what is at stake: not only whether students used a tool, but whether they are still developing the knowledge, judgment, and responsibility that higher education promises to cultivate.
Those concerns deserve better than dismissal, and they deserve better than detector dependence. They deserve assessment practices that produce evidence faculty can inspect, students can understand, and institutions can trust.
Key takeaways
- Faculty concern is less about the tool and more about whether students are still developing judgment and understanding.
- Polished work can rise in quality without a matching rise in learning, which is what unsettles assessment.
- The worry is broader than academic integrity: it spans student development, workplace readiness, and credential credibility.
- Connecting the artifact to the student’s own explanation addresses the concern without treating every student as a suspect.
References
College Board. (2026, February 27). College faculty weigh in on the role of AI in higher education. https://allaccess.collegeboard.org/college-faculty-weigh-role-ai-higher-education
D2L. (2025, June 11). New research data reveals that faculty workload can drop with daily AI use. https://www.d2l.com/newsroom/tyton_partners_report_examines_ai_in_higher_education/
Elon University. (2026, January 21). Elon/AAC&U national survey: 95% of college faculty fear student overreliance on AI. https://www.elon.edu/u/news/2026/01/21/elon-aacu-national-survey-95-of-college-faculty-fear-student-overreliance-on-ai/
Ruediger, D., Blankstein, M., & Love, S. (2024). Generative AI and postsecondary instructional practices: Findings from a national survey of instructors. Ithaka S+R. https://sr.ithaka.org/publications/generative-ai-and-postsecondary-instructional-practices/
Tyton Partners. (2025, October 30). GenAI and higher education: Safeguarding integrity and ensuring durability. https://tytonpartners.com/genai-and-higher-education/
Editorial note
This article was drafted with AI assistance and reviewed, revised, and approved by Provenance Learning’s human authors. The claims, sources, examples, and final wording were reviewed before publication. We believe responsible AI use should be transparent, reviewable, and subject to human judgment. See our Editorial Transparency statement.
Provenance is an artifact-based oral assessment platform for higher education. If you’d like to see a session run on one of your own assignments, request a demo.