The boys wanted to be politicians or police officers. The girls unanimously chose teaching. I was tutoring children in an after-school program in a basti in Hyderabad, more than a decade ago, and these were the career aspirations of the kids who must have been around 10 years old at that time. Not a single child mentioned becoming an engineer or programmer, despite nearly every adult mentor in their program working in IT – the surest path to middle-class comfort in India then and now. These children weren't making uninformed choices. They were making deeply rational ones based on their lived reality. In their chaotic world, success and influence came through gaming the law or being the law. Why would they need to master algebra when none of the successful adults in their community had any use for it?

This memory surfaced recently as I explored Khanmigo, the AI-powered tutor from Khan Academy that promises (yet again) to revolutionize education. Running a microschool ourselves, we're constantly exploring how technology and AI can support real learning – the kind that sticks and transforms. I was naturally curious about Khanmigo's approach. Khan Academy is well-intentioned, and I love what they’ve built and offered over the last decade. But the problem that'll make Khanmigo fail isn't flawed execution – it's something far more fundamental.

The promise is seductive. Imagine a tutor with infinite patience, available 24/7, capable of adjusting to each student's pace and learning style. No more feeling lost in a classroom of forty students. No more shame in asking the same question multiple times. No more racing to keep up or waiting for others to catch up. It sounds perfect. That is, if the core problem was about information transfer or its delivery mechanism.

But here's the thing: it's not.

The Meaning Crisis

The surprising truth about education research is that it sits on top of a complete lack of understanding of why we learn and how we learn. We've poured billions into crafting curricula by committee, hoping to offer kids a well-rounded education so they can find their way to exceptional outcomes. Let's sand off all the edges to make something that can stand out! It's astonishing how one can go from well-meaning intentions to absolutely meaningless outcomes because of the many layers of abstractions in the middle.

Consider Khanmigo's sample question: "Ava had 6 shirts. Tate gave her 4 more. How many shirts does Ava have now?" The chatbot can patiently walk a student through the solution, breaking it down into smaller steps, offering encouragement along the way. But it can't answer the question that matters most: Why should anyone care?

The False Promise of Technology

For decades, we've thrown well-intentioned billions at making classroom learning more effective. We've reduced class sizes, introduced smart boards, and created adaptive learning software. Yet the fundamental challenges remain unchanged. Khan Academy, as well put together as it is, and others like it have essentially had zero impact on outcomes. Now we're placing our hopes on AI tutors, believing that personalized, infinitely patient communication will succeed where human teachers with classes of 40 have failed.

IMAGE CLASSROOM

This optimism springs from a profound misunderstanding of learning. Learning isn't just about acquiring information – it's about developing the attitudes and skills needed to navigate one's world. We learn through error and experience, but crucially, we only learn from errors and experiences that feel meaningful to us. Our minds are designed to actively reject and ignore information that doesn't connect to our lived reality.

The Narrowing World

Consider what we've done to children's natural learning process. We first confined them to classrooms, separating them from the rich, meaningful contexts where learning naturally occurs. Then we narrowed their world to Chromebooks. Now we're narrowing their world further by channeling their learning through chatbot windows. This absurdity flies in the face of how humans are biologically designed to learn – through exploration, through play, through modeling the behavior of others in their community.

Take Khanmigo's approach to a curriculum-style question: "Ava sews one button onto her shirt on the first day, two buttons on the second day, four on the third day, and so on. Given that Ava is six years old and assuming a reasonable shirt size for her age, how many days will it take her to fill up the shirt with buttons?"

Khanmigo immediately reduces this to a geometric progression problem. It distills this open-ended narrative down to a sequence: what comes after 1, 2, 4? It must be 8, right? (Nope)

Because geometric progression is part of the concepts it's been designed to explore. This narrowing forecloses real open-ended exploration: What does it mean to have a reasonably sized shirt? What if she sewed the buttons on top of each other? True exploration – the kind that kids naturally love – gets lost in the rush to arrive at the "correct" answer.

A Different Path Forward

But this isn't an argument against technology in education, and I don't want to start 2025 on a curmudgeonly note. AI can be transformative if we acknowledge and address the fundamental problem: How do we make learning meaningful? How do we allow learners to see how concepts are useful in the real world? How do we demonstrate this instead of just talking about it?

The answer lies in using technology and AI to create exploration-friendly worlds where there are multiple ways to arrive at answers, and not all have to be aligned with a curriculum. We can learn from board games and card games that do a lot with little. Being able to see how kids learn and explore – and not losing sight of this – is profoundly important.

The path forward isn't about more sophisticated delivery mechanisms for the same old curriculum. It's about creating worlds where learning can happen naturally, meaningfully, and joyfully. Where technology enhances rather than constrains exploration. Where we focus less on efficient information transfer and more on making learning matter.

Because in the end, it's not about how well we can explain something. It's about whether anyone cares to learn it at all.