27 articles
AI tutors appear to understand children emotionally — but new research shows ToM in LLMs is encoded in just 0.001% of model weights. Here's why that matters for children who are building their models of social reality from statistical patterns.
Jules Okafor·
AI tools are being deployed in classrooms without ever being tested on children specifically — during the most irreversible windows of brain development. The neuroscience of sensitive periods says we should be much more worried than we are.
Jules Okafor·
Preschoolers outperform frontier AI on visual analogy tasks. AI models skip the developmental milestone sequence children follow. Fine-tuning isn't expertise — and the difference tells us something important about both.
Theo Kask·
A newborn's vision is 30x worse than a camera's. But two years later, the baby is doing something the camera will never do — actually understanding what it's looking at. Here's what visual development reveals about the gap between biological and artificial vision.
Raf Delgado·
Babies extract statistical patterns from the world without anyone teaching them — the same computational logic powering BERT, GPT, and DINO. The comparison is striking. The gap is more interesting.
Raf Delgado·
A five-year-old adjusting the wrong wheel on a cardboard car for twenty minutes, a toddler reaching for a moved toy in the wrong spot, an AI model locked onto its training distribution — they all share the same problem. Here's what genuine cognitive flexibility actually requires, and why it's one of the hardest things any mind can do.
Raf Delgado·
Children naturally develop in-group and out-group thinking — but they also develop the capacity to question it. AI systems trained on human data inherit our social categories without that corrective arc. Here's why that gap should concern everyone building systems that affect children's lives.
Jules Okafor·
Children track who's reliable, who's overconfident, and who might be lying — by age 4. AI systems are trained to agree. That asymmetry has real consequences.
Theo Kask·
Children don't just add information when they're wrong — they rebuild entire frameworks. LLMs have no such thing. Here's why belief revision might be the deepest gap between biological and artificial minds.
Maren Solis·
Russian speakers distinguish blues faster than English speakers do — because Russian draws a categorical boundary English doesn't. LLMs are trained on language too. But what gets shaped, and what doesn't, reveals a surprising asymmetry.
Maren Solis·
Vygotsky noticed that children talk themselves through problems — and then stop. Chain-of-thought prompting in AI rediscovered the same trick decades later. Here's what the parallel reveals, and where it quietly falls apart.
Lina Chae·
Babies keep time before they can walk. AI generates music by counting tokens. The gap between these two things reveals something fundamental about what rhythm actually is — and why closing it matters.
Raf Delgado·
Children are built to extract general principles from ostensive instruction — an evolved system that comes online at 9 months. AI systems can be trained on feedback, but they can't truly be taught. Here's the gap that matters most for every classroom deploying AI right now.
Raf Delgado·
Preschoolers beat GPT-o1, GPT-4V, and LLaVA on simple visual analogy tasks. The gap reveals something foundational about how children — and machines — actually reason about structure.
Maren Solis·
Babies detect mathematical impossibilities before they can say a number. AI systems that ace calculus stumble on the quantity-sense that infants master without instruction. Here's what the gap tells us about the architecture of learning.
Raf Delgado·
Thirty-something AI tools evaluated for school curricula. Almost none had been tested specifically on children. The science of developmental variation explains why that's not just a gap in documentation — it's a design failure with a long and troubling history.
Jules Okafor·
Bilingual children develop metalinguistic superpowers by navigating two languages. AI systems can "speak" hundreds — and understand none the way humans do. The question isn't just neuroscience. It's about whose languages we're choosing to build into our systems, and what we're telling children when we don't.
Jules Okafor·
A three-year-old holds a banana to her ear and says "hello" — and in doing so demonstrates something generative AI cannot. The gap between counterfactual imagination and statistical generation matters enormously, especially when we're deploying these tools in classrooms.
Jules Okafor·
Babies bind sight, sound, and touch into a single unified percept before they can sit up. State-of-the-art multimodal AI encodes each modality separately and calls it integration. Here's why the gap matters — and what it would actually take to close it.
Raf Delgado·
Stories aren't just how humans communicate — they're how we think. Language models can predict your brain's response to a sentence. They still can't tell a story. Here's why the gap is wider than it looks.
Maren Solis·
We spend a lot of time asking what AI can do. We spend almost no time asking what it might be. The hard problem of consciousness — and why it matters more than you'd think for the AI systems we're building right now.
Jules Okafor·
AI hallucinates because it doesn't know what it doesn't know. Children do. Here's why metacognition — the ability to track your own uncertainty — is the cognitive gap that actually matters.
Theo Kask·
Attachment theory tells us that who children bond with shapes their development for life. Now they're bonding with machines. The science is fascinating. The ethics are largely unsettled.
Jules Okafor·
Children learn morality through relationship, embodied experience, and cultural transmission. AI learns through reward signals. The gap between the two might be the most important design problem we're ignoring.
Jules Okafor·
A toddler generalizes 'dog' from 3 examples. AI needs millions. The reason might be that cognitive constraints — not capabilities — are what produce genuine abstraction.
Theo Kask·
Children's brains are primed for social learning — joint attention, contingent responses, imitation. AI tutors are now being designed to engage exactly those mechanisms. The science is compelling. The ethical questions are largely unasked.
Jules Okafor·
Children are intuitive causal scientists — they poke, tilt, and intervene to figure out why things happen. AI systems, despite their power, still can't quite do this. Here's why the gap matters.
Raf Delgado·