There is a growing assumption in education that artificial intelligence will finally deliver what decades of reform have not: personalized learning, improved outcomes, and a more efficient system. Classrooms are adopting AI at a breathtaking pace. Students are using it to write essays, solve problems, and complete assignments. Schools, eager not to fall behind, are integrating it into instruction with remarkable speed.
But beneath this enthusiasm lies a more uncomfortable truth:
Artificial intelligence is not entering a system designed for learning. It is entering a system designed for academic performance. Technology should not replace the critical relationship between teacher and student. That essential bond — built on empathy and trust— cannot be outsourced.
And that distinction may determine whether AI becomes the most powerful educational tool in history—or one of its greatest threats.
School Was Never About Learning
For all the rhetoric about “student-centered learning,” modern schooling has long been organized around a different objective: academic performance.
Grades, test scores, completion rates, and credentials have served as the primary signals of success. Students are rewarded for producing correct answers, not for developing durable knowledge. Instruction is often paced for coverage, not mastery. Assessment measures outputs, not understanding.
In this system, learning—the slow, effortful process of building knowledge in memory—has been secondary.
For over a century, this orientation has not simply emerged—it has been structured.
The expansion of standardized testing—shaped in part by institutions such as the Educational Testing Service—helped define achievement through measurable outputs (Lemann, 1999; OECD, 2023). At the same time, the Carnegie Foundation for the Advancement of Teaching formalized the Carnegie Unit, standardizing schooling around time—seat hours—rather than mastery (Tyack & Cuban, 1995; Carnegie Foundation, 1906/modern analyses).
Together, these forces helped build a system where time, the pace of curriculum coverage, and performance became proxies for learning.
This is not merely a historical observation. It is reflected in decades of research showing that instructional practices and teacher beliefs remain misaligned with how learning actually works (OECD, 2023; National Academies, 2018). Despite major advances in cognitive science, schools have not systematically integrated principles such as retrieval practice, spacing, and the role of prior knowledge into everyday teaching.
The result is a system that produces performance without mastery.
And nowhere is this more visible than in subjects like STEM, where each concept builds on prior knowledge. When foundational understanding is weak, later learning collapses. Students appear to progress, but their knowledge remains fragile, fragmented, and easily forgotten.
A System That Never Learned How to Teach Learning
If schools are not optimized for learning, part of the reason lies upstream—in how educators themselves are prepared.
Research consistently shows that teacher education programs have historically failed to incorporate the science of learning (Pomerance et al., 2016), and that this gap persists today, with limited and inconsistent coverage of evidence-based principles in teacher preparation (Surma et al., 2019; NCTQ, 2020–2025; Schmied & Jamaludin, 2023).
Teachers are often asked to improve student outcomes without access to a coherent, research-based understanding of how memory, practice, and prior knowledge shape learning (National Academies, 2020; OECD, 2023). Professional development, when it exists, is frequently fragmented and disconnected from classroom practice (Kraft & Hill, 2020).
The consequences extend beyond instruction to students themselves. Despite robust evidence that strategies like retrieval practice and spacing are far more effective, students continue to rely on rereading and highlighting—approaches that feel productive but yield weak long-term learning (Carpenter et al., 2022; Murray et al., 2025).
In other words, schools have not only struggled to teach content. They have struggled to teach students how to learn.
Enter AI: The Perfect Tool for the Wrong System
Into this already misaligned system comes artificial intelligence.
AI excels at producing outputs—answers, essays, solutions—with speed and precision. In a system that rewards outputs, this appears to be an advantage.
But research suggests a deeper problem.
Studies show that heavy reliance on AI is associated with reduced critical thinking, driven by cognitive offloading—the outsourcing of mental effort to external tools (Benedek & Sziklai, 2025). In experimental settings, students using AI demonstrate lower cognitive engagement than those working independently (Benedek & Sziklai, 2025). Other scholars warn that unstructured AI use may lead to “cognitive atrophy,” as learners bypass the effort required to build durable knowledge (Lodge et al., 2026).
At the same time, students using AI often produce higher-quality work while demonstrating weaker understanding of the underlying material (Adejumo, 2026). The OECD has similarly noted that AI can improve performance without producing sustained learning gains when it replaces core cognitive processes (OECD, 2026).
This is the central paradox of AI in education:
It can make students look more capable than they actually are.
From Learning Crisis to Inequality Crisis
If schooling were designed around learning, this paradox might be manageable. But in a system already optimized for performance, AI amplifies existing weaknesses.
Students who already possess strong knowledge and self-regulation can use AI to extend their thinking. Those who struggle are more likely to use it to bypass effort altogether (Yunus et al., 2025). The result is not just stagnation—it is widening inequality.
Over time, this dynamic risks creating a bifurcated system:
One group uses AI to enhance human intelligence
Another becomes dependent on AI in place of it
The implications extend beyond education.
A society in which knowledge resides increasingly in machines rather than in people is not merely an educational concern. It is a civic one.
If learning is not the goal of schooling, AI does not solve the problem. It accelerates it.
The Antidote: Reclaiming Learning in the Age of AI
The solution is not to reject AI. It is to redesign education around a principle that has too often been neglected:
Learning requires thinking. And thinking cannot be outsourced.
Research suggests that AI can support learning when used as a scaffold rather than a substitute. AI-based tools can enhance self-regulated learning—helping students plan, monitor, and reflect—when they are embedded within structured instructional routines (Adejumo, 2026; Frontiers in Education, 2025).
Students must be expected to attempt tasks independently before using AI. AI should be used to provide feedback, ask questions, and surface misconceptions—not to generate final answers. Reflection must become central: students should explain how their thinking changed, justify their decisions, and verify outputs against reliable sources (Wang et al., 2026).
Assessment must also evolve. If AI can produce polished outputs, outputs can no longer serve as proxies for learning. Schools must prioritize tasks that reveal thinking directly—retrieval, explanation, transfer, and application (OECD, 2026).
And educators must be equipped with a working understanding of how learning works. Without this foundation, even the most powerful technologies will be misapplied.
The Real Choice
The debate over AI in education is often framed as a choice between innovation and resistance. It is neither.
The real choice is whether we will continue to operate a system focused on performance—or finally build one centered on learning.
If we do the former, AI will make that system more efficient—and more unequal.
If we do the latter, AI may help us achieve something education has long promised but rarely delivered: not just better performance, but deeper, more durable learning.
Technology is not the determining factor.
The purpose of schooling is.
References
Carnegie Foundation for the Advancement of Teaching. (1906). The Carnegie Unit.
Lemann, N. (1999). The Big Test: The Secret History of the American Meritocracy. Farrar, Straus and Giroux.
National Academies of Sciences. (2018). How People Learn II.
OECD. (2023). Trends Shaping Education.
Tyack, D., & Cuban, L. (1995). Tinkering Toward Utopia. Harvard University Press.
A cynical teacher of mine (he taught economics) once told me that, “School exists to train you to sit for 8 hours a day doing something you don’t want to do. Passing exams means that you can demonstrate to a future employer that you are willing to apply yourself sufficiently to that 8-hour block of doing things that inherently don’t interest you to excel at it. In many respects, that demonstrates to them that you’ll make an ideal employee, but don’t mistake that for an education.”
Learning requires time and adaptation.. as an educator (20 years) I have seen how these two elements are constantly ignored.. “not keeping pace to an overloaded curriculum = bad teacher” 🤦🏻♀️