Key Takeaways
- Although 63 % of R‑1 universities actively encourage generative AI use, the empirical evidence on its learning benefits is mixed and often negative.
- Frequent reliance on AI correlates with weaker critical‑thinking skills and a phenomenon called “cognitive surrender,” where students relinquish agency to the machine.
- Long‑term knowledge retention suffers when AI is used as a study aid; short‑term performance gains can mask durable learning deficits.
- Durable learning depends on deep cognitive engagement, productive struggle, repetition, and meaningful human interaction—processes that AI can bypass if it simply supplies answers.
- Faculty should adopt a “learning‑first” default: design instruction around proven learning conditions and integrate AI only when it demonstrably supports, rather than substitutes for, those conditions.
- Four practical questions help decide whether an AI tool is appropriate: does it reinforce core knowledge, require transfer to new contexts, support independent reasoning, and preserve human interaction?
- AI developers’ profit motives mean enthusiasm for the technology should not be mistaken for pedagogical effectiveness; the burden of proof lies with the technology, not the skeptical instructor.
Introduction & Adoption of Generative AI in Higher Education
Colleges and universities nationwide are rapidly embracing artificial intelligence, driven by the promise that AI will sharpen student thinking, personalize learning, and better prepare graduates for a technology‑saturated workforce. An analysis of sixty‑five R1 institutions revealed that 63 % actively encourage the use of generative AI, with many publishing detailed guidance for its classroom integration (McDonald et al., 2025). While institutional enthusiasm is high, the underlying assumption—that AI inherently improves learning—requires careful scrutiny, as a growing body of research suggests a more complicated picture.
Evidence Against AI’s Learning Benefits
The core assumption that AI enhances learning is not consistently supported by empirical findings. A Swiss study reported a negative correlation between frequent AI tool use and critical thinking: the more students offloaded cognitive work to AI, the weaker their critical‑thinking abilities became (Gerlich, 2025). Wharton researchers identified a related phenomenon they termed “cognitive surrender,” wherein participants accepted AI‑generated outputs with little or no scrutiny, effectively transferring agency to the machine (Shaw & Nave, 2026). A 2025 meta‑analysis of eighteen generative‑AI studies confirmed that over‑reliance on AI undermines higher‑order thinking skills such as critical analysis and problem‑solving (Qu et al., 2025). These findings challenge the optimistic narrative surrounding AI adoption in academia.
Cognitive Surrender and the Loss of Agency
Cognitive surrender differs from deliberate cognitive offloading; it represents a wholesale relinquishment of mental effort to an AI system. When students allow AI to produce answers without engaging in evaluation or justification, they miss opportunities to develop independent judgment and analytical skills. This passive consumption of AI‑generated content can erode the very competencies that higher education aims to cultivate, leaving learners adept at using tools but deficient in the ability to think critically on their own.
Broader Technology Trends and Declining Cognitive Performance
Concerns about AI are part of a larger pattern observed since the early 2000s, when K–12 schools began mass‑adopting laptops and tablets. International assessments such as PISA, TIMSS, and PIRLS have documented declining performance that correlates with heavier technology use (Horvath, 2026; Rogelberg, 2026). Notably, IQ scores have fallen in ways without historical precedent, suggesting that pervasive digital tools may be impairing foundational cognitive capacities. While AI is the latest iteration of this trend, its potential impact on learning must be examined within this broader context.
Randomized Controlled Trial on ChatGPT and Knowledge Retention
A 2025 randomized controlled trial—the gold standard of educational research—provides particularly compelling evidence against uncritical AI use. Students who employed ChatGPT as a study aid retained significantly less knowledge 45 days after instruction compared with peers who studied without the AI tool (Barcaui, 2025). Although short‑term performance (e.g., immediate quiz scores) sometimes showed gains, these benefits masked long‑term learning deficits. The study underscores that AI can create an illusion of mastery while impairing durable knowledge acquisition.
How AI Shortcuts the Learning Process
Durable learning hinges on deep cognitive engagement, productive struggle, and repetition—processes that AI can circumvent when it simply supplies summaries, essays, or problem solutions. For example, asking education students to analyze an AI‑generated lesson plan requires less cognitive work than having them write one themselves. The latter activity forces learners to wrestle with fundamental questions: How do I write a clear learning objective? How do I structure this activity for diverse learners? The resulting brainstorming, drafting, revising, and justification constitute the learning itself. Using AI as a shortcut is akin to taking steroids instead of training at the gym; the effort that builds strength is bypassed, leaving only superficial results.
A Learning‑First Default and the Four‑Question Framework
Given the research, faculty should adopt a default of “offline pedagogy”—designing instruction around conditions known to produce durable learning—and integrate AI only when it can be shown to genuinely support those conditions. Those conditions include building a strong content knowledge base, providing opportunities for deep processing and productive struggle, fostering independent and critical thinking, and preserving meaningful human interaction. To operationalize this approach, instructors can work through four questions before integrating any AI tool:
- Will this AI tool help students use, recall, and demonstrate understanding of core disciplinary content? Higher‑order thinking rests on domain knowledge; AI should engage students with foundational material rather than let them bypass it.
- Will this AI tool require students to apply their learning to a new context? Transfer of knowledge to novel situations signals genuine understanding; AI should scaffold, not perform, that transfer.
- Will this AI tool support—not replace—independent, evidence‑based reasoning? Critical thinking demands that students justify their decisions; after using AI, they must be able to articulate their reasoning in their own words.
- Will this AI integration preserve meaningful human interaction? Peer feedback, collaborative problem‑solving, and instructor‑student dialogue develop social and intellectual habits that AI cannot replicate; any tool that diminishes these interactions risks more harm than benefit.
If an AI tool fails to meet these criteria, the prudent default is to leave it out, placing the burden of proof on the technology rather than on the skeptical educator.
Proceed with Caution: Balancing Innovation and Evidence
AI is unlikely to disappear from higher education, and blanket resistance would be unrealistic. Nevertheless, the pace of adoption is outstripping the accumulation of rigorous evidence. Faculty often feel pressure to implement tools endorsed by their institutions and already familiar to students, without clear guidance on whether such use will help or harm learning. It is also essential to remember that AI developers are motivated by profit, not pedagogical efficacy; enthusiasm from technology companies should not be conflated with proof of educational value. The four‑question framework offers a principled starting point for decision‑making, ensuring that AI serves as a supplement to, rather than a substitute for, the cognitive processes that underpin lasting learning. By anchoring innovation in proven learning principles, educators can harness AI’s potential while safeguarding the intellectual development of their students.