You were not born curious or incurious. You were born with a dopaminergic system perfectly designed to become obsessed with unsolved problems. Everything that happened after that either exercised that system or let it atrophy. This week: 3 papers that together explain what curiosity actually is, whether it can be trained, and why it may be the cognitive capacity that matters most in the age of AI.

Robot-Proof has launched and so much going on (see the media round-up below) I'm trying to restart my "Mad Science Solves.." AMA series, but for now come join my for the Kennedy Human Rights Center Book Club on April 6 (11 Pacific/2 Eastern). Sign up here!

Research Roundup

The Curiosity Chemical

I don’t usually share general media articles in my newsletter, but this BBC opinion piece, “What we get wrong about dopamine”, makes a claim worth attention: dopamine doesn't make you feel good; it makes you obsessed.

That's the correction that neuroscience has been trying to make for years, and it has implications that run well beyond the pop-psychology conversation about reward and motivation.

The standard account—dopamine as the brain's pleasure chemical, the neurochemical of satisfaction and enjoyment—is not just imprecise. It's backwards. Dopamine isn't released because you got the reward. It's released in anticipation of a reward that is not yet certain, and its function is not to register pleasure but to strengthen the neural pathways that produced the successful behavior. Wherever dopamine is released, memories consolidate more deeply. The brain is receiving a signal: do more of what just led here.

This is why Adderall (which works by driving available dopamine out of dopamine-producing neurons) creates focus and tunnel vision but not euphoria. The drug isn't making you feel good. It's making you care intensely about completing the task in front of you. Rats injected with amphetamine don't show facial expressions associated with pleasure, but they work harder for rewards, with no increase in apparent enjoyment. Dopamine is not the chemical of getting what you want. It is the chemical of needing to figure out how to get it.

The most striking demonstration in this review comes from a set of experiments with pigeons that has since been replicated across species, including humans. Pigeons given a button that delivers rewards at predictable intervals—50 pecks per reward, 100 pecks per reward—show fatigue and reluctance as the ratio increases. They're performing a transaction, and the terms are getting worse. But make the reward interval unpredictable—1% chance of reward on each peck—and the pigeons don't stop even though the total expected reward is the same. They continue pecking obsessively, far past the point where a rational cost-benefit analysis would suggest stopping. What drives them isn't the reward. It's the unsolved pattern. It's the not-yet-knowing.

Even weirder: deliver rewards at completely random intervals, with no button at all, and most pigeons will eventually start pecking a button. They invent the pattern when there is no pattern to find. The dopaminergic drive to resolve uncertainty is so powerful that the brain will manufacture agency rather than tolerate randomness.

This is what curiosity is neurochemically. Not a personality trait. Not a gift some children have and others don't. It’s a biological drive—the dopaminergic system's response to unresolved uncertainty—that pulls the organism toward the unknown with the same compulsive force that it pulls an addicted brain toward its substance. The difference is what the uncertainty is about, and whether the environment has been structured to reward the resolution of genuine intellectual unknowns or to provide the hit without requiring the work.

The implications for everything downstream—how we design learning environments, what social media and AI are doing to our curiosity, why some people become more intellectually alive with age while others become less so—follow directly from this mechanism. I'll be returning to them for the rest of this issue and in Thursday's paid newsletter.

Dopamine is not a pleasure chemical (endogenous opioids handle that…its right there in the name). It is the “figure this out” chemical. And if you want to understand why some minds become addicted to ideas versus empty sensation, you need to understand what that drive requires in order to stay alive.

Curiosity Can Be Trained

If dopamine is the neurochemical of unresolved uncertainty, if curiosity is the brain's drive to crack the pattern, then it follows that curiosity is not a fixed trait but a practiced capacity. Like any capacity that has a biological substrate, it can be exercised or left to atrophy. The question is what the exercise looks like and how early it has to start.

A preregistered experiment with 103 children between 5-7 years old gives a precise answer to the first part of that question and a quietly radical answer to the second.

The experimental design was straightforward [1]. Children were split into two groups for one-on-one science lessons over two weeks. One group was encouraged to listen carefully. The other was encouraged to ask questions. That was it; attention rewards for listening vs questioning. No special curriculum, no elaborate training protocol, no technology. Just a consistent framing of what the child's job was during the lesson: to receive or to inquire.

Children in the question-asking condition valued new science information significantly more than children in the listening condition. They didn't just perform differently on tests of what they'd learned. Their relationship to learning itself had shifted. They wanted more.

It’s worth particular attention that children with less domain knowledge—lower baseline vocabulary and science achievement—showed greater benefits from question-asking practice than their more knowledgeable peers.

This inverts the standard educational logic, which tends to teach the facts first then ask the questions later, once students have enough to work with. These findings suggests giving the child the experience of inquiry first, before the domain knowledge consolidates, and the knowledge that follows lands in a brain that has been share by the inquiry. The dopaminergic system is already predicting reward for deeper inquiry. The information arrives as an answer to a question the child has already asked, rather than as a fact to be filed.

There is a methodological caveat I want to name directly. Two weeks is a short intervention, and an effect size of Wilcoxon r = 0.23, while meaningful for something this small, needs replication at longer time scales to tell us anything about durability. The researchers note this themselves. What makes me take the finding seriously despite the brevity is the preregistered design—the hypothesis was committed to before the data was collected—and the mechanistic coherence with everything the dopamine literature predicts.

The practical implication is both hopeful and uncomfortable. Hopeful because the intervention required to shift children's relationship to inquiry is genuinely accessible: it costs nothing, requires no technology, and can be implemented in a single lesson by a single teacher who simply changes the question she asks the child. Uncomfortable because the dominant structure of formal education does almost exactly the opposite: it rewards careful listening, penalizes wrong answers, and systematically removes the uncertainty that the dopaminergic system requires to stay engaged.

We have been training the wrong capacity. And we have been doing it to children young enough that the training sticks.

Knowledge doesn't kill curiosity

It sharpens it into something the machines can't replicate.

There is an assumption embedded in a great deal of educational anxiety (and in a great deal of the AI conversation) that expertise and curiosity are in tension. That knowing the answers crowds out the drive to ask the questions. If the truly curious mind is the beginner's mind, then accumulating domain knowledge is, at some level, a trade-off against the open, exploratory cognitive posture that generated the curiosity in the first place.

A longitudinal study of undergraduate psychology students says this assumption is half right, and the half that's wrong is the more important half.

Researchers measured students' question-asking abilities at the beginning and end of a semester-long introductory course, both general question-asking ability and domain-specific question-asking ability. They also tracked performance on two very different assessments:

1. a closed-ended multiple-choice exam and
2. an open-ended final project.

Domain-specific question-asking improved significantly over the semester, while general question-asking ability remained stable or actually declined. Knowledge doesn't suppress curiosity. It redirects it from broad, exploratory, low-resolution wondering toward precise, targeted, high-stakes inquiry. The beginner asks "why does this happen?" The expert asks "under what conditions does this mechanism break down, and what would the exception tell us about the rule?"

Most fascinating to me: general and domain-specific question-asking abilities were negatively related to closed-ended test performance but positively related to open-ended project performance.

The students who were best at asking questions performed worse on the multiple-choice exam; yet they performed significantly better on the complex, open-ended project.

I want to be honest about my ambivalence on the first half of this finding. I'm not entirely sure what the negative relationship between question-asking and closed-ended test performance is telling us, and I don't want to overclaim. One interpretation: the question-asking mind has already moved up the cognitive hierarchy—it's building models, exploring implications, attending to edge cases—and perhaps it’s less efficient at the specific retrieval operation that a multiple-choice test rewards. The highly curious student, having genuinely engaged with the material, may have organized it in ways that are richer but less test-optimized than the student who memorized the answers. That account feels plausible but uncertain.

The second half of the finding is golden: open-ended, complex, ill-posed project performance—precisely the performance type that cannot be reduced to retrieval and cannot be easily substituted by a tool—is driven by the capacity to ask the right questions, not to know the right answers.

This is what domain knowledge actually buys you, and why the framing of expertise as the enemy of curiosity is wrong. You cannot ask the precise, high-leverage question that cuts to the heart of an unsolved problem without enough domain knowledge to know where the heart is. General curiosity gives you the drive. Domain knowledge gives you the aim.

Here is where this connects directly to the AI moment, and to the argument I make in Robot-Proof: closed-ended test performance—the retrieval of known answers to well-posed questions—is exactly what AI does well, much better than any human alive. Open-ended project performance—the navigation of ill-posed problems filled with genuine uncertainty toward novel insights—is what AI cannot do without a human who can ask the questions that frame the problem.

It’s also why “knowing things” is still valuable in a world where AI agents know everything. While the students who were best at asking questions performed worse on the test that measures what AI can replace, they performed better on the task that measures what it cannot. Knowledge gave structure to inquiry.

If you want to be robot-proof, you need the foundational knowledge required to ask the highly specific, domain-relevant questions that machines simply don't know how to ask. Not because the knowledge itself is irreplaceable. Because knowledge is what turns a general dopaminergic drive toward uncertainty into a precision instrument pointed at the problems that actually matter.

Media Mentions

Fast Company: Everyone keeps saying reskilling will save us from AI displacement. I've spent years studying the data. It won't—the reason why goes back to a post-war myth most people have gotten completely wrong.

My latest in Fast Company: "The Reskilling Delusion" — an excerpt from Robot-Proof, out now. The argument isn't that workers can't learn. It's that we're training for the wrong thing entirely.

Private Company Directors: Boards are under pressure to "adopt AI" — and most are doing it in exactly the way that will hurt them.

I wrote about the efficiency trap: why optimizing for AI-driven cost reduction may be the most dangerous strategic mistake a board can make right now. In Private Company Directors: "AI, Boards, and the Efficiency Trap"

Deborah Kalb: Finally, enjoy this conversation with Deborah Kalb what it takes to be "robot-proof".

Much, much more to come!

SciFi, Fantasy, & Me

Susanna Clarke's Piranesi is the most precise fictional portrait of curiosity I know—not curiosity as a means to an end but as a complete orientation toward the world. Its protagonist inhabits an inexplicable house of infinite halls and tides and statues, and responds not with fear or frustration but with patient, systematic, joyful inquiry: cataloguing, hypothesizing, revising, seeking the pattern whether or not a pattern exists. Reading it is a reminder of how much we lose when we train curiosity out of people. (Of course, the protagonist looses other things in the process.)

Stage & Screen

• March 17, Online: The book launch! Robot-Proof: When Machines Have All The Answers, Build Better People is will finally be inflicted on the world.
• March 27, Geneva: Its a secrect, but I'll tell you soon.
• March 30, Amsterdam: What else: AI and human I--together is better!
• April 2, Paris: A book reading on the Seine!
• April 14, Seattle: Ill be keynoting at the AACSB Business School Conference.
• April 16, NYC: A private event in Brooklyn. The topic is AI, but I'll make it about us.
• May 12, Online: I'll be reading from Robot-Proof for the The Library Speakers Consortium.
• May 12, SF: We'll talk about collective intelligence, the neuroscience of trust, and how dumb I have to be to be launching my 13th company.
• May 14, Miami: TEDxMiami
• June 9-10, London: London Tech Week!
• June 11, Luxembourg: How Europe (and even some of it smallest states) compete and grow in a trade environment dominated by zero-sum leaders
• June 12, Denver: GlobalMindEd
• June 18, Stockholm: The Smartest Thing on the Planet: Hybrid Intelligence
• October, Toronto: The Future of Work...in the Future

Vivienne L'Ecuyer Ming