Two pieces of news caught my attention this week.

The first was the passing of Lee Shulman, a giant in educational research, whose profound understanding of teaching and learning shaped generations of educators – including myself. The second was the approval of a new virtual charter school in Arizona that proudly declares it will have no teachers. None. Zero. Just AI.

The timing couldn’t be more ironic. The gulf between these two visions of education couldn’t be wider.

Shulman understood that “Learning is least useful when it is private and hidden; it is most powerful when it becomes public and communal.” Yet here we have a school built on the opposite premise – learning through private interactions with a black box AI system. No teachers. Just “brand consultants” for motivation and emotional support.

Kids attending these schools start their day with two hours of instruction on topics led by the Unbound’s purportedly specialized AI program, which can allegedly adjust the learning plan as it goes to tailor to how a child is doing on each subject…. After their AI inculcation session, the kids are let loose to work on extracurricular tasks where they can apply the wisdom they gleaned from a machine learning model.

I have no words…

The contrast, of course, is stark.

Shulman’s vision emphasized how knowledge grows stronger when shared within a community, tested and examined by peers and mentors. Real learning happens in the dynamic space of the classroom – where teachers guide and probe, where they listen and explain, then listen some more. They adapt their teaching based on the look on a student’s face, or the gleam of understanding. When needed, they find another way to explain the same concept, drawing on their pedagogical content knowledge to make ideas accessible. It’s a deeply human process of constant adjustment and response.

Walk into any vibrant classroom and what do you hear? The buzz of conversation, the back and forth of questions and answers, the excited sharing of ideas. Some might call it noise. But that’s not noise – that’s the sound of learning being made public. That’s the sound of understanding being constructed, of ideas being tested and refined through dialogue. Listen carefully and you’ll hear students building on each other’s thoughts, challenging assumptions, seeking clarification. You’ll hear a teacher moving between groups, asking just the right question at just the right moment to push thinking further. This is what Shulman meant by learning becoming “public and communal” – it’s dynamic, messy, and gloriously human.

Of course, one might argue that this ideal of teaching and learning doesn’t always happen in our schools. That many classrooms fall short of this vision of dynamic, communal learning. That’s a fair criticism. But in a real classroom, there’s always the possibility – the hope – that such moments of shared understanding will emerge. That’s the power of having humans in the room, of maintaining the space for these interactions.

In an AI school? That possibility is designed out of the system from the start. There is no room for the serendipitous moment of collective insight, no space for the unexpected question that transforms everyone’s understanding. The very design of the system precludes it.

It is as if we have just given up!

And of course, they pray at the altar of efficiency – proudly advertising teaching at “twice the pace” of traditional classrooms. As if speed were the point of education. As if learning were simply about moving through a sequence of steps derived from averaging countless other such sequences.

And we haven’t even addressed the more obvious concerns – the well-documented problems with AI hallucinations and biases, the massive environmental costs of training these models, the privacy implications of children’s data. It is not as if this technology is ready for prime time—far from it, as this list of the biggest AI fails of 2024, shows. But these technical issues, while serious, mask the deeper problem. At its core, this represents a fundamental shift toward command and control of learning – all couched in the seductive language of efficiency and personalization. All in the guise of helping kids.

This approach to education creates multiple layers of distance – physical distance between learner and teacher, emotional distance in the learning process, and conceptual distance from what we know about how learning actually works.

This reminds me of what I have previously written about as the “reductive seduction of other people’s problems” – the further we are from an issue, the easier it seems to solve.

But education is messier than that. More complex. More human. More deeply embedded in the contexts where learning happens.

I often wonder about those who propose educational solutions for other people’s children. And to be clear, this is not a new pattern. Tech leaders have a history of protecting their own children from the very technologies they promote to others. As I noted in a previous blog post, Chamath Palihapitiya, former Facebook VP of Growth said of the product he was charged with promoting, his children were “not allowed to use that s**t.”

Perhaps the simplest measure of any educational innovation is this: Would you want your own child to experience it? And if the answer is no, then take a breath and ask why? As I think about a school proudly declaring “We won’t have any academic teachers,” I wonder if its proponents would send their own children there?

My guess is their children would not be allowed near that s**t!