Three pendulums hang from a string line: one with paperclips you can add to make it heavier; one with tape marks so you can change the length; and a small bin of paperclips with a sign that reads, “Save these for later—today we only change the length.” “No lecture today,” she says. “Write your prediction. Give it a confidence number. Then test.”
The room settles into problem-solving noise—chair squeaks, pencil taps, the whisper-fast shuffle of index cards. Henry and Kira burn two swings changing mass and wince. “We’re burning tries,” Kira mutters, sketching a little grid so they won’t repeat themselves. Across the room, Susan squints at her notes, resets the mass, and shortens the string by a finger width. She watches once, twice, then looks up with the certainty of a dog rounding a corner.
“We can stop now,” she grins. “It’s length, not mass.”
A chatbot could spit out the pendulum law instantly. But with a clock ticking and a rule you can’t break, these kids learned something rarer: how to shrink uncertainty—together, in real time.
Eight careful tries beat one borrowed answer.
Parents can practice this at the breakfast table. Ask: “Do bubbles pop faster on the cold glass or the warm one?” Make a guess, give it a confidence number, run one quick test, stop. Learn. Homework: “Which two cuts make this paragraph clearer?” Mark, read aloud, decide, done.
One more thing for the big moments: tell your kids not to use AI to write their essay or college application. Have them write it themselves—voice, choices, fingerprints—and then use AI like a coach to tighten the draft. Ask the model why it suggests each change. Keep the authorship; borrow the diagnostics.
Let AI supply examples. You coach the habit of testing and stopping. Machines can be thorough. Humans must be decisive.
Teachers know this rhythm, too: shorten the lecture, lengthen the “cheap test.” Five minutes of setup, then loops of predict → try → check. A lab group suspects friction, lowers its confidence, adjusts—and a week later transfers the insight from strings to springs. That’s the skill worth exporting: the thing you earn today that works again tomorrow.
Facts are on tap; judgment is the bottleneck.
Scientists sketch curiosity in everyday terms. There’s interest curiosity—the tug toward something new—and deprivation curiosity—the itch to close a gap. Curiosity isn’t a mood; it’s the brain’s control system for learning. When something truly surprises us, a small burst of dopamine says “learn this,” tightening focus and pulling us toward the next test. The sweet spot is a Goldilocks zone—too easy and we’re bored; too hard and we quit; just right and attention locks on.
Information is plentiful; discernment is scarce.
Species express it through their strengths. Dogs follow scent and detours. Dolphins ping the dark with sound, adjusting every second. Humans predict, test, and—when wise—stop in time.
Today’s AI doesn’t itch. It doesn’t nose the wind. It produces the answers we request but never the next question.
If we tested intelligence the curiosity way, the exam would be simple: eight tries, predictions with confidence, a rule flipped midstream, then a new problem with the same hidden logic.