Back from Frankfurt—Where the Hippocampus People Met the Basal Ganglia People

Cheery Friday Greetings from Barb Oakley!

Back from Frankfurt—Where the Hippocampus People Met the Basal Ganglia People

Last month I promised to report back from Frankfurt, where Terry Sejnowski and I co-chaired the Ernst Strüngmann Forum on what offloading our thinking onto AI does to the brain. It was dazzling, especially as helmed by the imitable Julia Lupp. Some forty researchers spent a week in four working groups whose reports will become a book. The best part was getting people who ordinarily never talk to each other into the same room. Neuroscientists who work on the deep mechanics of memory almost never meet the edtech people who watch what students do with ChatGPT. Even inside neuroscience, the hippocampus people hang out with the hippocampus people, the basal ganglia people with the basal ganglia people, and the poor cerebellum is everybody’s stepchild. Here, everybody got to meet everybody—the sharp wave ripples of our mind-wandering helping us to become even more creative.

The same goes for cognitive psychologists, even the visionary ones. They simply don’t often have a chance to talk to in-depth researchers in genAI, or the in-depth neuroscience that bears directly on their own work, and vice versa. Let me give you an example from my background paper. Researchers at Anthropic recently performed a kind of surgery on a large language model. They removed a thin sliver of its internal activity, less than a tenth of the whole. The model still parsed text, recalled facts, and spoke in polished, confident prose. The trouble appeared when the model had to reason through several steps. (Anyone who has listened to a glib student smoothly confabulating to cover for missing knowledge will recognize the phenomenon.) When the researchers fed the model long lists of unrelated words, only about six fit in that sliver at once. Nelson Cowan, the working-memory researcher who was with us in Frankfurt, puts the human focus of attention at three to five chunks. Now, were Anthropic researchers just inadvertently mimicking known models of the human brain with their findings? Or does this relate at a foundational level to the limitations of the three-dimensional world (or four-dimensional world, if you count time).

The Forum’s book will take a while, but one distinction was clear. Neuroscientifically speaking, offloading after you have learned something is very different from offloading instead of learning it. A calendar offloads memory without doing your thinking for you. Generative AI can do the selecting, organizing, and reasoning before a learner has done any of it himself. Picture a student who gets stuck, asks the AI, receives a polished answer, and feels relief. Now picture the same student who first retrieves what he can, sketches a poor attempt of his own, and only then asks the AI to critique it. He ends up unstuck either way. But the brain strengthens whatever action produced the relief, so the first sequence trains him to outsource while the second trains him to retrieve.

Who is most at risk here? One point I can’t overemphasize is that youngsters learn very differently from adults. Part of the reason is that the brain’s two learning systems—the declarative system that builds conscious understanding and the procedural system that turns practice into automatic skill—mature, level off, and decline at different rates. The same AI shortcut that saves a professional an hour can cost  fourteen-year-olds the mental framework, or schema, that they were supposed to be building. I’ll report more once the book is in the offing.

GOTO Copenhagen, September 29–30

Don’t forget, on Tuesday, September 29, I’m teaching a full-day masterclass at GOTO Copenhagen, “Learning How to Learn in an Age of AI,” for anyone who wants to learn better in a world of AI—teachers, parents, managers, and developers alike. We’ll dig into cutting-edge research on critical thinking and AI.  (It’s very rare for me to teach a full day masterclass—GOTO is pretty special—so take advantage of the opportunity if you can!) The next day, Wednesday, September 30, I’ll be keynoting the conference with “Grokking the Brain: What AI Reveals About How You Really Learn”—how AI’s strange habit of suddenly “grokking” a pattern after endless memorization gives us our clearest picture yet of how your own brain builds expertise.Details and registration at https://gotocph.com/2026.

Now in Spanish: Acelera tu aprendizaje con ChatGPT

My Coursera course with Vanderbilt computer scientist Jules White, Accelerate Your Learning with ChatGPT, now has an entirely Spanish sibling, Acelera tu aprendizaje con ChatGPT, which launched on September 8. The team at Universidad Francisco Marroquín in Guatemala, where I hold an honorary doctorate, produced the course, with UFM’s awesome Joana Paglia and Daniel Mazariegos as the instructors. In about four hours it covers how your brain learns, how to use ChatGPT to extend your working memory and to get personalized practice and feedback, and why memorization and critical thinking still matter in the age of AI. If you know someone who learns in Spanish, please pass the course along.

researchED Toronto, October 24—One Month Out

researchED Toronto is a month away, with a terrific lineup. The keynotes are Nick Gibb, the former UK schools minister behind England’s reading turnaround; Kareem Weaver, the Oakland, California, literacy activist who has spent years insisting that learning to read is a civil right; and Zaretta Hammond, who connects the neuroscience of learning to culturally responsive teaching. Patricia DeGuire, chief commissioner of the Ontario Human Rights Commission, which produced the Right to Read inquiry into how Ontario’s schools teach reading, gives the opening remarks. Anita Archer, a longtime authority on explicit instruction, is presenting. My own session works one math example all the way down to the brain—the different jobs of declarative and procedural memory, what scaffolding physically consists of, and how JUMP Math teaches division. The conference runs Saturday, October 24, at Timothy Eaton Memorial Church, with an opening event on Friday evening, October 23. Registration is on Eventbrite (register soon, because researchEds sell out fast!). If you’ll be there, come find me!

Critical Thinking Makes the Best-Of List

Class Central, the site that catalogs more than 250,000 online courses, just published its guide to the best critical thinking courses out there—and I was tickled to see that my course, Critical Thinking: A Brain-Based Guide for the ChatGPT Era, made the cut. If you’ve been meaning to tune up your reasoning for an era when AI can sound utterly sure of itself and still be wrong, consider this your nudge. Check it out! Best Critical Thinking Courses Guide.

That’s all for now. Happy learning!

Barb Oakley

• Learning How to Learn—one of the world’s most popular MOOCs, with its recommended text, A Mind for Numbers 
• Uncommon Sense Teaching—the book and Coursera Specialization!
• Mindshift—the book and MOOC
• Learn Like a Pro—the book and MOOC
• For kids and parents: Learning How to Learn (for Youth)—the book and MOOC. Pro tip—watch the videos and read the book together with your child. Learning how to learn at an early age will change their life!