On AI and pompoms

A slice of life: in the past few weeks, I’ve given a dissertation proposal talk, published a journal paper, co-organized a 200-person community learning festival, attended an AI conference full of tech founders and policy leaders, and run a hands-on workshop for 60 teenagers about screen addiction — where a central activity involved decorating paper phone cases with pompoms, pipe cleaners, and stickers.

I keep thinking about the pompoms and phone cases.

The workshop was at a museum, and the teenagers arrived with the usual mix of rebelliousness, opinions, and low-key phone dependency you’d expect. The premise was simple: excessive phone use is a dopamine-linked addictive behavior, not just a personal failing. That’s a research finding. But you can’t hand a research finding to a 15-year-old and expect it to change anything. What you can do is put craft supplies on a table and ask them to design something — to identify their own habits and triggers, translate what they’ve learned into a tangible object, and start thinking about how the design of a thing changes our awareness of it and relationship to it. How design — and thus an opportunity for self-action — sits in the space between what we want and what we “want to want”. By the end, they were genuinely engaged in a design problem; they were scaling up from designing a tangible object to thinking about the design of bigger systems and strategies. Not just because they’d absorbed the research, but because they were doing something with it.

The next afternoon, I was at the conference. The conversations were about human agency, frontier models, distributed data, policy frameworks. I felt privileged to have conversations with brilliant researchers and builders — but when asked about my own work, I found myself hesitating before mentioning the previous night’s phone addiction workshop. It seemed comically handmade, too low-fidelity, far from policy-level. I talked about my empirical research instead — my RCTs looking at perception of human and AI presence, the learning qualities emerging from different types of interactions. Conference-appropriate language for conference-appropriate people.

But the hands-on building, even with pompoms, feels just as important to me. I keep coming back to this: the phone case workshop and the empirical studies are asking the same question. How do you help people develop real capability — not just absorb information, but actually change how they think and act? The studies give me evidence about what matters (interaction quality over content delivery, genuine effort over self-reported satisfaction). The workshop is where I test whether I can design for those things with actual humans in a room. And the research would be worse without the workshop, because watching teenagers navigate a design problem teaches me things about motivation and engagement that no quantitative survey can.

I think the instinct to separate these — to present the research at the conference and keep the craft supplies out of sight — is the same instinct that makes a lot of AI-in-education work feel disconnected. We talk about capability development in the abstract while often the lived examples of it are happening in rooms with craft supplies and teenagers who would likely rather (at want-level) have been scrolling at home. The altitude at which we discuss these questions isn’t always the altitude at which they get answered.

Jane Jacobs has a line that I love: that cities "have the capability of providing something for everybody, only because, and only when, they are created by everybody." She was making a point about scale. Cities need to function as enormous systems, but they're actually built and experienced at the level of an intersection, a stoop, a pile of leaves. I think the same is true of learning. The big questions about AI and human capability will get answered partly in policy rooms and research labs, and partly in rooms where a teenager is gluing pompoms to a paper phone case and accidentally discovering something about how design shapes behavior.