Colleges and universities across the country are moving quickly to embrace artificial intelligence. According to an analysis of sixty-five R1 institutions, 63 percent of them actively encourage the use of generative AI, with many publishing detailed guidance for its classroom integration (McDonald et al., 2025). The implicit promise is that AI will sharpen student thinking, personalize learning, and better prepare graduates for a technology-saturated workforce.
But a growing body of research tells a more complicated story—one that faculty, instructional designers, and academic leaders should not ignore.
The Evidence Problem
The assumption driving AI adoption is that it improves learning. Yet the evidence, at best, is inconsistent—and in many cases, points in the opposite direction.
A Swiss study found a negative correlation between frequent AI tool use and critical thinking: the more students offloaded cognitive work to AI, the weaker their critical thinking became (Gerlich, 2025). Wharton researchers went further. Across multiple experiments, participants accepted AI-generated outputs with little or no scrutiny—a phenomenon they termed “cognitive surrender.” Unlike deliberate cognitive offloading, cognitive surrender involves a wholesale transfer of agency to the machine (Shaw & Nave, 2026). These findings are not isolated: a 2025 meta-analysis of eighteen generative AI studies confirmed that over-reliance on AI tools undermines higher-order thinking skills, including critical analysis and problem-solving (Qu et al., 2025).
The concern extends beyond AI specifically. Since K–12 schools began adopting laptops and tablets en masse in the early 2000s, IQ scores have fallen in ways that have no historical precedent. International assessments—PISA, TIMSS, and PIRLS—show declining performance correlated with heavier technology use (Horvath, 2026; Rogelberg, 2026).
Perhaps most striking is a 2025 randomized controlled trial—the gold standard of educational research—in which students who used ChatGPT as a study aid retained significantly less knowledge 45 days after instruction than students who studied without it (Barcaui, 2025). Short-term performance gains masked long-term learning deficits.
Why AI Shortcuts the Learning Process
Durable learning requires deep cognitive engagement, productive struggle, and repetition. When students use AI to summarize a reading, draft an essay, or guide them through a problem, they may produce polished outputs—but they are not building lasting knowledge or skill. The shortcut bypasses the very processes that create learning.
Consider a teacher education course. Is it more valuable to ask students to analyze an AI-generated lesson plan, or to write one themselves? The latter demands that students wrestle with real questions: How do I write a clear learning objective? How do I structure this activity for diverse learners? That productive struggle—brainstorming, drafting, revising, justifying—is the learning. AI-generated scaffolding, however well-intentioned, can short-circuit it entirely. As one analogy goes: it’s like taking steroids instead of training at the gym. The whole point is the effort.
This does not mean AI has no legitimate role in higher education. It means that role must be defined carefully, with student learning—not engagement, efficiency, or career readiness—as the primary criterion.
A Learning-First Default
Given what the research tells us, faculty should adopt a default of “offline pedagogy”—designing instruction around the conditions known to produce durable learning—and integrate AI only when it can be shown to genuinely support, rather than substitute for, those conditions. Those conditions include:
Building and reinforcing a strong content knowledge base; Opportunities for deep processing and productive struggle; Independent and critical thinking; and Meaningful human interaction.
The following four questions translate these conditions into a practical decision-making framework. Before integrating AI into any lecture, activity, or assignment, faculty should work through each one.
Four Questions for Deciding Whether to Use AI
Question 1: Will this AI tool help students use, recall, and demonstrate understanding of core disciplinary content?
Higher-order thinking is built on a foundation of domain knowledge. Students cannot analyze a lesson plan without understanding what learning objectives and assessments are. They cannot evaluate a scientific argument without knowing the relevant concepts. If an AI tool actively engages students with foundational content—through retrieval practice, targeted feedback, or elaborative questioning—it may be worth integrating. If it simply allows students to bypass that content, it is almost certainly counterproductive.
Question 2: Will this AI tool require students to apply their learning to a new context?
Transfer—applying knowledge to a novel situation—is one of the most reliable indicators of genuine understanding. When a student applies principles of lesson design to a new grade level or subject, they are moving information from temporary working memory into more stable long-term knowledge. If an AI tool scaffolds that transfer while preserving cognitive effort, it can be valuable. If it performs the transfer for the student, learning is short-circuited.
Question 3: Will this AI tool support—not replace—independent, evidence-based reasoning?
Critical thinking requires students to make judgments and defend them. A student writing a lesson plan must decide how to open the lesson, how to group students, and how to assess understanding—and then justify those decisions with pedagogical reasoning. Any AI integration that substitutes the AI’s judgment for the student’s own undermines this process. The test is simple: after completing the task, can the student articulate—in their own words—why they made the decisions they made?
Question 4: Will this AI integration preserve meaningful human interaction?
Peer feedback, collaborative problem-solving, and instructor-to-student (and student-to-student) dialogue do more than support academic learning—they develop the social and intellectual habits that define educated citizens. Human interaction sparks curiosity, broadens perspective, builds trust, and provides the kind of accountability that AI cannot replicate. Before integrating any AI tool, ask whether it complements or competes with these interactions. An AI-enhanced discussion board that replaces peer response with algorithmic feedback may sacrifice more than it gains.
Proceed with Caution
AI is not going away, and blanket resistance is neither realistic nor necessarily wise. There are genuine use cases where AI can support learning without undermining it. But the pace of adoption in higher education is far outrunning the pace of evidence. Faculty are often caught in the middle—pressured to integrate tools their institutions endorse and their students already use, without clear guidance on whether doing so will help or harm the people they are trying to educate.
It is also worth remembering that AI developers have profit motives that have nothing to do with improving student learning. The enthusiasm of technology companies should not be mistaken for evidence of pedagogical effectiveness.
The four questions above will not resolve every instructional dilemma, but they provide a principled starting point. If an AI tool cannot clearly support content knowledge, productive struggle, independent reasoning, and human interaction—the conditions we know produce learning—the default should be to leave it out. The burden of proof belongs to the technology, not to the faculty member who questions it.
Norman Eng, EdD, is a lecturer at the School of Education, Brooklyn College (CUNY) and founder of EducationXDesign, Inc., which provides instructional training and workshops to higher education faculty. Find out more at NormanEng.org.
References
Barcaui, A. (2025). ChatGPT as a cognitive crutch: Evidence from a randomized controlled trial on knowledge retention. Social Sciences & Humanities Open, 12, 1-13. https://doi.org/10.1016/j.ssaho.2025.102287
Gerlich, M. (2025). AI tools in society: Impacts on cognitive offloading and the future of critical thinking. Societies, 15(1), 6. https://doi.org/10.3390/soc15010006
Horvath, J. C. (in press). The digital delusion: How classroom technology harms our kids’ learning—and how to help them thrive again. Penguin Random House.
McDonald, N., Johri, A., Ali, A., & Hingle Collier, A. (2025). Generative artificial intelligence in higher education: Evidence from an analysis of institutional policies and guidelines. Computers in Human Behavior: Artificial Humans, 3, 100121.
Qu, X., Sherwood, J., Liu, P., & Aleisa, N. (2025). Generative AI tools in higher education: A meta-analysis of cognitive impact. In Extended Abstracts of the CHI Conference on Human Factors in Computing Systems (CHI EA ’25), Yokohama, Japan. ACM. https://doi.org/10.1145/3706599.3719841
Rogelberg, S. (2026, February 21). The U.S. spends $30 billion to ditch textbooks for laptops and tablets: The result is the first generation less cognitively capable than their parents. Fortune. https://fortune.com/2026/02/21/laptops-tablets-schools-gen-z-less-cognitively-capable-parents-first-time-cellphone-bans-standardized-test-scores/
Shaw, S. D., & Nave, G. (2026). Thinking fast, slow, and artificial: How AI is reshaping human reasoning and the rise of cognitive surrender. Working paper, The Wharton School, University of Pennsylvania. http://dx.doi.org/10.2139/ssrn.6097646
