AI in the Classroom: What Changes, What Shouldn't

When calculators appeared in classrooms in the 1970s and 80s, a vocal segment of educators predicted the collapse of mathematical understanding. Students would become dependent on machines, lose the ability to think mathematically, and graduate unable to perform basic arithmetic.

Fifty years later: calculators are standard educational tools, mathematics instruction has evolved rather than collapsed, and nobody seriously argues we should go back. The question was never whether to use calculators — it was how.

AI in education is the calculator debate at a much larger scale. And right now, we're in the panicky early phase where the discourse is dominated by extremes: AI will destroy education (cheating! dependency! the end of learning!) versus AI will save education (personalized tutoring for every child! no more one-size-fits-all!). Both positions contain truth. Neither is complete.

The Personalization Promise

The single most powerful argument for AI in education is personalization. In a typical classroom, one teacher manages 30-40 students who learn at different speeds, have different strengths, struggle with different concepts, and respond to different teaching styles. The teacher can differentiate somewhat, but the fundamental constraint is that one person cannot simultaneously provide thirty personalized learning experiences.

AI tutoring systems can. Not perfectly — we're early — but directionally. A system like Khan Academy's Khanmigo can identify that a student consistently struggles with fractions involving unlike denominators, and provide targeted practice on that specific gap rather than making the student redo an entire fractions unit. It can notice that a student learns better from visual explanations and shift its approach accordingly.

From my observation, the impact is most significant for students at the extremes — those who are far behind the class average and those who are far ahead. Both groups are poorly served by class-paced instruction. The struggling student is lost and embarrassed. The advanced student is bored and disengaged. An AI tutor can meet both where they are, simultaneously, without making either feel singled out.

The Cheating Problem Is Real But Overblown

Yes, students use ChatGPT to write essays. Yes, this is a problem. No, it's not the existential crisis some educators describe.

Students have always found ways to avoid doing their own work. Pre-AI shortcuts included: copying from classmates, paraphrasing Wikipedia, buying essays from paper mills, and the time-honored tradition of "my older sibling did the same assignment three years ago." AI makes cheating easier and the output higher-quality, but it doesn't create the motivation to cheat — that comes from assignments that feel meaningless, grading systems that reward output over understanding, and educational cultures that prioritize compliance over curiosity.

The more productive response — which some educators are already implementing — is to redesign assessment rather than simply banning AI. Oral examinations. Process-based grading (where the thinking process matters more than the final product). In-class writing with AI discussion afterward. Assignments that require personal experience or local knowledge that AI can't fabricate convincingly.

An essay prompt like "Analyze the themes in Macbeth" is trivially outsourced to AI. A prompt like "Interview someone over 70 about a decision they regret, and connect their experience to a theme in Macbeth" is not. The second is also a better assignment, AI concerns aside.

What Teachers Tell Me

I've talked to about a dozen teachers in Indian schools and colleges about their experience with AI in their classrooms. The responses cluster into three groups:

The fearful (about 30%): Worried about cheating, concerned about their own relevance, some have tried to ban AI tools entirely. Understandable anxiety, but the ban approach rarely works — students use AI on their phones regardless of policy.

The cautious adopters (about 50%): Using AI selectively — as a lesson planning assistant, as a tool for generating practice questions, occasionally as a classroom aid. These teachers tend to have the most nuanced views: they see both the benefits and the risks, and they're experimenting carefully.

The enthusiasts (about 20%): Fully integrating AI into their teaching, encouraging students to use it as a learning tool, redesigning their courses around AI collaboration rather than AI avoidance. These teachers consistently report that student engagement has increased, though they note it requires significantly more preparation time on their end.

What Should NOT Change

Some things in education are irreplaceable by AI, and I think it's important to name them explicitly because the enthusiasm for AI-everything can obscure what matters most:

The relationship between teacher and student. A good teacher doesn't just transfer information — they model curiosity, demonstrate how an expert thinks, notice when a student is struggling emotionally rather than intellectually, and provide the kind of encouragement that comes from knowing a student as a person. AI can supplement this. It cannot replace it. The students I remember as transformative in my education were teachers, not textbooks. No amount of personalized AI tutoring replicates what a caring adult with expertise provides.

Learning to struggle. The process of being stuck on a problem, trying multiple approaches, failing, and eventually finding a solution (or learning to ask for help) is genuinely educational. AI that provides instant answers short-circuits this process. Students need to learn that confusion is part of learning, not a bug to be immediately resolved.

Social learning. Debating ideas with peers, explaining concepts to someone who doesn't understand them (which deepens your own understanding), collaborating on group projects — these are skills that AI cannot teach because they require other humans. Education that optimizes for individual AI tutoring at the expense of group learning trades short-term efficiency for long-term social and intellectual development.

The calculator didn't kill math education. AI won't kill education broadly. But the transition will be bumpy, the debates will be heated, and the teachers who navigate it best will be those who understand both what AI can do and what only a human can.