970x125
Lately, the role of artificial intelligence (AI) in education seems split down the middle. AI is good, or AI is bad. But I’ll argue that this argument is being framed too simply. One side celebrates speed and efficiency. The other warns that students are outsourcing thought and losing the friction that makes learning real. Both sides are seeing something important. Neither is getting to the center of it.
What matters isn’t merely whether AI is present in education—like an on/off switch that shines a light but fails to illuminate. What really matters is where it sits inside the learning process.
That distinction has become more important to me as I think about two very different observations. In one case, I wrote about how cognitive performance can fall off quickly when AI support is removed, even after a short period of use. In another, I looked at the Nigerian study suggesting that students completed a two-year curriculum in about six weeks with AI-supported instruction.
At first glance, those ideas seem to tug against each other. How can AI dramatically accelerate learning while also threatening the integrity of learning itself? I don’t think that is a contradiction. I think it reveals that AI can do two very different things in education, depending on how it is used. It can support learning, or it can hollow it out.
Faster Is Not the Same as Deeper
Education often collapses very different outcomes into a single word. We say “learning” when we may actually mean performance.
A student can quickly complete more problems and produce better-looking work. Of course, that matters, but it’s not the whole story. Growth runs deeper. Growth is what happens when knowledge begins to “settle into the self” and becomes usable without the external push of AI.
AI is very good at improving performance, and in the right setting, that can be enormously helpful. But human development has always depended on more than fluency. It depends on the cognitive bumps of uncertainty and correction. Far from a flaw to be corrected by AI, it’s the mechanism by which understanding becomes durably human.
AI Can Strengthen or Weaken Learning
The Nigerian study is both fascinating and important. But it shouldn’t be oversimplified.
The intervention wasn’t just a chatbot dropped in front of students. It was structured. It involved human supervision, the measurement of educational goals, and an academic framework. AI wasn’t replacing the curriculum; it was being used within one.
That’s very different from a student depending on AI as a convenient engine of completion. In that setting, the machine does the work, from solving math questions to writing essays. The outputs can improve, and the student may appear more capable. But something essential may be diminished or eliminated completely. The student may be finishing tasks without fully building the cognitive competence those tasks were meant to develop.
So, yes, AI can accelerate what many call learning. But it can also crush learning into something more procedural and less developmental.
The Real Question of Placement
Maybe it’s time to stop treating AI in education like an on/off switch. The question is not simply whether students should use it or whether schools should ban it. That frame is already too crude.
AI has to be crafted into the curriculum.
If AI enters the learning process too early, it can interrupt the productive struggle that builds minds. If it enters later, after a student has been introduced to the material, it can play a very different role. It can take on a unique iterative dynamic that drives a newfound learner centricity.
A well-designed and forward-thinking classroom might ask students to draft an argument before using AI to critique it. It might require a first attempt at solving a problem before AI is allowed to offer alternatives or explanation. It might use AI to create more individualized feedback after the student has already done the first layer of cognitive work.
That’s the scaffold. That’s Lev Vygotsky’s zone of proximal development.
But when AI does the first thinking, when it supplies the structure before the student has had to build one, education begins to slide toward substitution. And once that becomes a habit (for student and teacher), the damage may not be obvious at first. Students can still sound fluent, and they can still turn in polished work. But fluency is not mastery, and polish is not proof of understanding.
What Kind of Learner Are We Building?
This is why the debate feels so unsatisfying. Too much of it swings between enthusiasm and alarm, and I might be guilty of this. One side sees access and acceleration. The other sees dependency and cognitive decline. Both are pointing to something real.
The harder truth is that AI is not one thing educationally. Its effect depends on the relationship it creates between the student and the work. If the tool helps a learner persist through difficulty while preserving the burden of thought, that is a gain. If it removes the burden too early, it may produce the appearance of progress while weakening the deeper machinery of growth.
Curriculum isn’t just a delivery system for information. It shapes habits of mind. It teaches students what to do when they do not know. It forms their relationship to effort, uncertainty, memory, and self-trust.
Once AI enters that space, it begins shaping those things, too.
That is why AI in education needs design, not devotion. The real question is not whether it belongs in the classroom. The real question is whether we are placing it in a way that still allows the student to become someone through the work. Because in the end, education is not just about reaching the answer. It is about fostering the kind of person who can still think when the technological augmentation is gone.
