Adaptability

TLDR: Adaptability is knowing how to learn—and knowing when to unlearn. In an AI age where knowledge becomes obsolete faster than ever, it’s the difference between children who leverage change and those who are overwhelmed by it.

Related Doing capacities: Agency, Persistence

Your child is studying. They’ve re-read the chapter three times, highlighted the key passages, quizzed themselves on the same problems in the same order. They feel ready.

They’re not.

Decades of cognitive science research shows that the study strategies that feel most effective are often the least effective. Re-reading creates a “fluency illusion”—a false sense of mastery that evaporates on test day. Blocked practice (same problem type, over and over) feels productive but produces shallow learning. Students using these methods believe they’re learning more when they’re actually learning less.

The strategies that actually work—interleaving topics, spacing practice over time, testing yourself instead of reviewing—feel harder and slower. Students often perform worse during training with these methods, then dramatically outperform on later tests.

Robert Bjork stumbled into this paradox while researching memory at UCLA. Experiment after experiment, he discovered something that frustrated educators: the conditions that produced the best immediate performance often produced the worst long-term learning. Students hated the methods that worked. They preferred the comfortable strategies that didn’t. Even after seeing the data, they kept choosing wrong.

Bjork called this the “illusion of mastery” and spent five decades proving our intuitions about learning are systematically backwards. The struggle is the signal.

This matters more than ever. In a world where knowledge becomes obsolete faster than any previous generation experienced, the ability to acquire knowledge matters less than the ability to update it. Children will need to unlearn as much as they learn.

And that requires a capacity most schools never teach.

What Adaptability Actually Is

Adaptability is knowing how to learn—and knowing when what you’ve learned no longer applies.

Here’s a way to see it: some kids hit a wall and try the same approach harder. Other kids hit the same wall and think, “That’s not working—let me try something different.” That shift—recognizing when to persist and when to pivot—is adaptability.

It has four pieces:

• Meta-learning — Understanding how you learn. Which strategies work for you? How do you know when you’re actually learning versus just feeling like you’re learning? When are you fooling yourself?

• Learning agility — How quickly can you pick up something new? This isn’t raw intelligence—it’s the efficiency with which you acquire new knowledge and skills when you need them.

• Unlearning — The capacity to let go. Outdated mental models, beliefs that no longer serve, habits that once worked but don’t anymore. This is often harder than learning something new.

• Transfer — Applying what you’ve learned in one context to novel situations. Can you take a principle from one domain and recognize where it applies elsewhere?

What adaptability isn’t

These get conflated, but they’re different:

“Isn’t this just being curious?” Curiosity is the desire to explore. Adaptability is the speed and efficiency of learning. A curious child might love exploring but struggle to master new domains quickly. An adaptable child can rapidly acquire whatever the moment demands—whether or not it initially sparks interest.

“What about growth mindset?” Growth mindset is the belief that you can learn. Adaptability is actually learning quickly. Meta-analyses show the correlation between growth mindset and achievement is modest (d = 0.05-0.14). Believing you can learn is necessary but insufficient.

“Isn’t adaptability just intelligence?” Intelligence is capacity. Adaptability is how you use and develop that capacity. The strategies you use, your metacognitive awareness, your willingness to unlearn—these all modulate what raw intelligence delivers.

The triad

Agency gets you off the starting line. Persistence keeps you going when it’s hard. Adaptability helps you change course when needed. Three capacities that work together: start → continue → adjust.

Here’s the tension: Persistence says keep going. Adaptability says try something different. Navigating when difficulty is productive struggle (persist) versus a signal your approach isn’t working (pivot) requires Judgment.

The key insight

Most children execute strategies. Adaptable children observe themselves executing—and change course when it’s not working.

That metacognitive loop separates efficient learners from those who work hard but spin their wheels.

The AI Complication

Each wave of technology changes what we need to learn. Writing meant we didn’t need to memorize everything. Search engines meant we didn’t need to remember where to find things.

AI is different. It can learn faster than you can. The question becomes: what happens to human learning when machines out-learn us?

AI makes some learning obsolete

Why memorize when you can look up? Why master calculation when AI calculates? The danger: children stop developing the foundational knowledge structures that enable higher-order thinking.

Transfer requires something to transfer from. If children outsource too much foundational learning, they may lack the schemas that enable rapid acquisition of new domains. The building blocks matter even when the building is done by machines.

AI creates dependency before competence

A 2025 study warns that AI tools may undermine children’s metacognition and critical thinking. The mechanism: if AI does your thinking, you don’t develop the capacity for independent thought.

Each use strengthens the pathway that says I need help with this. The spiral accelerates: less capacity leads to more outsourcing leads to less capacity.

AI can learn anything—but can’t tell you what’s worth learning

In a world of infinite information, the bottleneck shifts from accessing knowledge to selecting it. Children need the meta-learning capacity to identify what they should learn, monitor whether they’re actually learning it, and know when to change course.

AI can answer any question. Knowing which questions to ask is irreducibly human.

The neuroscience of why this matters

The prediction error problem. Learning is driven by dopamine signals that respond to the gap between what you expected and what happened. Research shows dopamine neurons fire when outcomes surprise us, strengthening the neural pathways that led to the prediction. When AI provides correct answers, there’s no prediction to be wrong about. The learning signal doesn’t fire.

The struggle signal. Bjork’s desirable difficulties work precisely because they create struggle. The effort required to retrieve information, to discriminate between interleaved problems, to bridge spaced intervals—this signals to the brain that the information matters. AI that eliminates struggle may eliminate the signal that consolidates learning.

The metacognitive development window. Metacognition develops progressively through childhood, with critical periods for different components. Children who outsource learning management to AI during these windows may not develop the neural architecture for independent self-regulation.

Surprising Finding: AI Can Either Build or Destroy Meta-Learning

The emerging research is clear: AI’s effect on adaptability depends entirely on design. AI tutoring systems that require students to struggle, explain their thinking, and identify their own errors can enhance metacognition. Systems that simply provide answers may atrophy it. A 2024 study found AI-generated materials enhanced student confidence but had inconclusive effects on actual learning. The comfortable path isn’t always the learning path.

Screen time beyond AI

Passive consumption vs. active learning. The brain learns through engagement, struggle, and feedback. Scrolling and watching don’t activate learning systems. A child watching educational videos may feel like they’re learning while gaining far less than one who actively retrieves and applies the material.

The attention fragmentation problem. Cognitive flexibility requires focused attention. Rapid context-switching encouraged by digital environments may produce pseudo-flexibility that’s actually fragmented attention. True flexibility means smoothly shifting when appropriate—not being unable to sustain focus on anything.

The practical question: Does this activity require my child to struggle, retrieve, and adapt? Or does it smooth the path and provide answers?

The Research: What We Know

The evidence base for learning how to learn is substantial—and the effect sizes are large.

Self-regulated learning strategies have major effects. A meta-analysis found learning strategies produced a d = 0.859 effect on academic achievement. That’s huge. Another meta-analysis of learning scaffolding found g = 0.500. Learning how to learn may matter as much as raw ability.

Metacognitive interventions work. A 2024 meta-analysis of 67 studies found metacognitive interventions had moderate effects on self-regulated learning and achievement (g = 0.48). Interventions delivered by teachers were more effective than researcher-administered ones—suggesting these skills can be integrated into regular practice.

The “desirable difficulties” are quantifiable. Multiple meta-analyses have established the effect sizes:

These aren’t marginal effects. The testing effect alone (g = 0.70) means students who test themselves score about 40% higher than those who simply restudy.

Physical exercise improves cognitive flexibility. A meta-analysis found moderate-intensity exercise produced large effects on cognitive flexibility (SMD = 0.86). Moving the body helps the brain adapt.

Surprising Finding: The Brain Can’t Tell Good Learning from Bad

The neural systems that produce the feeling of learning are disconnected from those that produce actual learning. Research on fluency illusions shows that when material flows easily, we feel we’ve mastered it—but easy processing predicts poor retention. The strategies that produce durable learning (interleaving, spacing, retrieval) feel worse but work better. Children can’t trust their gut about whether they’re learning.

These findings replicate across ages, cultures, and domains. The science of learning how to learn is mature.

Early Childhood (0-5)

What we know

Young children are learning machines. Their brains exhibit maximum neuroplasticity. They acquire language, motor skills, and social knowledge at rates adults cannot match.

But this doesn’t mean they’re adaptable in the metacognitive sense. Research shows children between ages 3-5 begin exhibiting early forms of metacognition—they develop “theory of mind,” understanding that others think differently. They talk themselves through problems. But sophisticated monitoring of their own learning is limited by prefrontal cortex immaturity.

Sensitive periods are at their peak. Language, phonemic awareness, social-emotional development—these windows are widest now.

The danger: Over-scaffolding prevents self-discovery. If adults always provide the answer, children don’t develop the experience of figuring things out themselves. The struggle that builds learning-how-to-learn gets outsourced before it begins.

What you can do

• The Prediction Game. Before reading a book, ask what they think will happen. Before an activity, ask what they think it will be like. Afterward, compare predictions to reality.

  Instead of: “Let’s read this book!” Try: “What do you think this story is about? Let’s find out if you’re right.”

  This builds the prediction-error loop that drives learning—and teaches children to notice when their expectations miss.

• The Struggle Wait. When children face difficulty, resist the rescue. Let them struggle. Offer hints rather than answers.

  Instead of: Solving the puzzle for them when they get frustrated. Try: “Hmm, what haven’t you tried yet?” And then silence.

  The temptation to help comes from your discomfort, not their need.

• The Private Speech Prompt. Young children naturally talk themselves through problems. Encourage this by modeling: “First I’m going to try this… hm, that didn’t work… let me try something else.”

• Multi-Domain Exposure. Music, movement, language, spatial play. Diverse contexts build cognitive flexibility by requiring the brain to adapt to different rule systems.

• The Error Celebration. When children make mistakes, respond with curiosity: “Interesting! What made you think that?” Errors are information. Treating them that way builds the habit of analyzing rather than hiding mistakes.

Surprising Finding: Students Actively Choose Worse Methods—Even When Told the Truth

Research shows students prefer blocked practice over interleaved practice even after being shown interleaving produces better results. In Bjork’s studies, participants who experienced both methods still rated blocked practice as more effective—despite scoring lower on tests. The feeling of fluency is so powerful it overrides data and personal experience. This pattern starts early. Telling children about effective strategies isn’t enough.

Middle Childhood (6-11)

What we know

This is when metacognitive capacity takes off. Research shows by ages 8-10, children demonstrate advanced abilities—assessing their confidence, employing strategic approaches, recognizing when they don’t understand.

Cognitive flexibility shows sharp increases between ages 7-9, becoming largely mature by age 10. This is the window for teaching flexible thinking—shifting strategies, considering alternatives, updating beliefs.

The danger: Study habits form during this period—often implicitly. Children may develop ineffective strategies (re-reading, cramming, blocked practice) that feel good but don’t work. These habits, once formed, are hard to change.

Research on fluency illusions shows even elementary students misjudge their learning based on ease of processing. They think smooth means learned. It doesn’t.

What you can do

• The Study Strategy Switch. Explicitly teach that strategies that feel easy are often less effective than strategies that feel hard. Name the paradox: “If it feels too easy, you might not actually be learning.”

• The Flashcard Protocol. Teach them to make their own flashcards, shuffle them (interleaving), space practice over days, and test themselves rather than just reviewing.

  Instead of: “Read over your notes again.” Try: “Close your notes. What do you remember? Now check what you missed.”

  The retrieval is the learning, not the review.

• The Confidence Check. After studying, before testing, ask: “How well do you think you know this? Rate it 1-10.” Then check. Over time, they learn to recognize when their confidence is misleading.

• The Transfer Question. After learning something, ask: “Where else could you use this? What’s this similar to?” Build the habit of looking for connections across domains.

• The Strategy Menu. Create a visible list of learning strategies. Help them select appropriate ones for different tasks. Vocabulary learning is different from math procedures is different from conceptual understanding.

Surprising Finding: Testing Yourself Beats Re-Reading by 40%

Adesope et al.’s meta-analysis of 118 studies found retrieval practice produces a g = 0.70 effect size compared to restudying. That’s equivalent to moving from the 50th to 76th percentile. Yet most students spend study time re-reading notes. The strategy that feels like learning (smooth, comfortable) produces weaker memories than the strategy that feels like testing (effortful, uncertain).

Adolescence (12+)

What we know

Adolescence brings the capacity for abstract reasoning and hypothetical thinking. Teenagers can think about thinking itself—analyzing how their mind works, recognizing biases, reasoning about possibilities.

The prefrontal cortex continues developing into the mid-twenties. The capacity for sophisticated metacognition—analyzing thought processes, recognizing biases, applying logical reasoning to abstractions—emerges fully during this period.

Transfer becomes possible at new levels. Adolescents can draw analogies across domains, extract abstract principles, apply learning to genuinely novel situations. But this capacity requires cultivation.

The danger: Fixed identity prevents adaptation. Adolescents develop strong self-concepts (“I’m not a math person,” “Languages don’t come naturally to me”) that become self-fulfilling. These identity-level beliefs can persist into adulthood.

What you can do

• The Learning Log. Have them keep brief records of what they studied, how they studied, and how they performed. Patterns emerge: “When I spread practice out, I do better on tests.” Self-generated insights beat advice.

• The Belief Audit. Periodically examine beliefs about learning: “I’m bad at memorizing.” “I’m not creative.” Challenge these with evidence. Where did the belief come from? Is it actually true?

• The Unlearning Practice. When new information contradicts old beliefs, make the conflict explicit.

  Try: “I used to think X. Now I’m learning Y. Why might I have been wrong?”

  This builds the capacity for belief revision—one of the hardest and most important aspects of adaptability.

• The Self-Test Habit. Before reviewing notes or asking for help, attempt recall first. “What do I actually remember about this?” The effort of retrieval—even when it fails—strengthens memory.

• The AI Prompt Practice. Teach them to use AI as a thinking partner, not an answer machine.

  Instead of: “Solve this problem for me.” Try: “Help me think through where I’m stuck. What questions should I be asking?”

  This preserves metacognitive engagement while leveraging AI capability.

At Any Age

Practices that apply across development:

• Model Your Own Learning. Let children see you learn—including struggle, confusion, strategy-switching, and eventual mastery.

  Instead of: Learning new things privately and presenting competence. Try: “I’m trying to figure this out. It’s confusing. Let me try reading it more slowly… Oh, I think I see it now—it’s like…”

  Your visible struggle normalizes the learning process.

• Name the Strategies. Don’t just use effective strategies—label them. “We’re interleaving because mixing topics helps your brain tell them apart.” “We’re spacing because forgetting a little actually helps remembering.” Explicit naming builds transferable knowledge.

• Distinguish Fluency from Learning. Regularly point out the paradox: “This feels easy, which might mean you need more challenge” and “This feels hard, which is actually a good sign.”

• Celebrate Pivots. “You figured out a new approach when the first one didn’t work—that’s adaptability” matters more than “You got the right answer.” Reinforce the process that produces outcomes.

Surprising Finding: Forgetting Is Part of Learning—Not the Enemy

The spacing effect works partly because you forget between sessions. Research shows that returning to material after a delay, the effort of re-retrieval strengthens memory traces more than continuous study. Studies on children’s naps confirm: preschoolers who nap after learning show better retention than those who stay awake. The forgetting forces the brain to work harder, which consolidates learning.

Special Considerations

Neurodivergent children

ADHD: Metacognition as compensation. Children with ADHD often struggle with executive functions underlying self-regulated learning—planning, monitoring, impulse control. But research shows explicit metacognitive training can improve these functions. Computer-based cognitive training improved both symptoms and cognitive functions. Physical exercise significantly improved working memory and flexibility. ADHD children may benefit more from explicit instruction because they can’t rely on implicit self-regulation.

Autism: Flexibility requires explicit teaching. Cognitive flexibility is often impaired in autistic children, manifesting as rigid behaviors and difficulty with transitions. But research shows computerized training can improve flexibility, with gains maintained at follow-up. Physical activity improved all three dimensions of executive function. Bilingual autistic children outperformed monolingual peers on flexibility tasks—the brain remains plastic.

Anxiety: Metacognition can help or hurt. Research shows anxious children often have heightened self-monitoring—too aware of their thinking, leading to rumination. But metacognitive therapy targeting maladaptive beliefs about thinking can reduce anxiety and improve academic persistence. The key is adaptive metacognition—not just awareness, but effective strategies for managing thoughts.

Gender differences

Studies find female students report higher usage of self-regulated learning strategies. Research on early self-control found girls exhibited higher levels in kindergarten, predicting superior reading achievement years later.

Research on metacognitive beliefs found females underestimate their capabilities—but this “conservative strategy” was associated with enhanced metacognitive efficiency. Girls may be more accurate at knowing what they don’t know.

The practical implication: Boys may need more explicit instruction in self-regulation. Girls may need encouragement to trust their capabilities, especially in domains (math, science) where social messages undermine confidence.

Where Things Go Wrong

The fluency trap

When learning feels easy, children stop. The fluency illusion tells them they’ve mastered material when they’ve merely been exposed to it. Research shows students prefer blocked practice even when interleaving produces better results.

The mechanism: easy processing feels like competence. But easy processing often means the material is already known, or isn’t being deeply encoded, or will quickly fade. The struggle signals—confusion, effort, error—feel like failure but are actually the signature of learning.

The expertise trap

Research shows successful learners can become rigid precisely because their methods have worked. When environments change, expertise becomes a liability. Strong mental models enable rapid pattern recognition—but blind you to patterns that don’t fit.

Children develop mini-expertise traps early. The child who succeeds in math by memorizing procedures resists conceptual understanding because the procedures work. Until they don’t—and then they lack the foundation to adapt.

The outsourcing spiral

Each time children use technology to do their thinking—AI to write, calculator to compute, search engine to answer—they strengthen the dependency pathway and weaken the capacity pathway. Less capacity leads to more outsourcing leads to less capacity.

The solution isn’t avoiding technology but using it strategically. Use AI to check thinking after you’ve thought. Use the calculator after you’ve estimated. Use search after you’ve retrieved what you remember.

The identity lock

By adolescence, many develop fixed identities around learning. “I’m not a math person.” “I have a bad memory.” These become self-fulfilling prophecies.

The antidote is evidence: actual experiences of learning in domains you “can’t” learn. Small wins that contradict the identity begin to shift it.

Adaptability without persistence

Just as persistence without adaptability is stubbornness, adaptability without persistence is flightiness—jumping ship at the first difficulty rather than pushing through productive struggle.

Children need to develop judgment about when difficulty signals “keep going” versus “try something different.” This discernment develops through experience with both persisting and pivoting—and feedback on when each was the right call.

The Research: Going Deeper

That’s the practical guide. What follows is for the curious—the neuroscience, the debates, the frontier. If you’ve got what you need, skip to Resources.

The neuroscience of learning

The prefrontal cortex is the brain’s executive control center—planning, decision-making, working memory, and cognitive flexibility. Flexibility skills begin developing in early childhood with sharp increases between ages 7-9, becoming largely mature by 10 but continuing to improve through adolescence.

The PFC develops through use. Children who regularly encounter situations requiring mental shifting—adapting strategies, considering perspectives, updating beliefs—strengthen these pathways. Predictable environments that don’t challenge may not build the same capacity.

Dopamine and prediction errors drive learning. When outcomes differ from expectations—when you’re surprised—dopamine neurons fire, strengthening synaptic connections. This explains why struggle and surprise are essential: they generate the neurochemical signal that consolidates memory.

The metacognitive network involves the anterior prefrontal cortex, supporting the ability to observe your own performance. This emerges progressively through childhood, with significant advances between ages 8-10.

Surprising Finding: Experts Learn Slower (Sometimes)

Research on expertise traps reveals that deep expertise can actually slow adaptation when domains shift. Experts have strong mental models enabling rapid pattern recognition—but these same models blind them to novel patterns. Children may sometimes have an advantage: less established knowledge means less to unlearn.

Where experts disagree

Transfer pessimism vs. optimism. For decades, cognitive scientists were pessimistic. Classic studies showed training one skill rarely transfers to others. But recent research is nuanced: transfer of specific skills is limited, but transfer of metacognitive strategies appears more robust. Teaching students to monitor learning seems to transfer better than teaching specific content.

The 10,000 hours debate. Ericsson’s deliberate practice research emphasized expert performance requires ~10,000 hours of focused practice. But subsequent research shows practice explains only about 30% of success. How you practice matters as much as how long.

Growth mindset effects are smaller than popularized. A 2022 meta-analysis found interventions have small effects (d = 0.05), becoming nonsignificant after adjusting for publication bias. Believing you can learn isn’t the same as knowing how to learn.

The AI tutoring question. Will AI personalization help or hurt self-directed learning? Early evidence shows AI tutoring can improve outcomes. But children who never learn to structure their own learning may develop dependency. The answer is probably design-dependent: AI that scaffolds metacognition helps; AI that does metacognition for the child harms.

The frontier

Proactive interference and unlearning. Research on how prior learning blocks new learning is increasingly relevant. When old knowledge conflicts with new, the old can actively inhibit acquisition of the new. As AI accelerates knowledge obsolescence, understanding how to help children update mental models becomes critical.

Computational models of learning. Researchers are building systems that learn how to learn. These may eventually inform educational design by identifying optimal sequencing, spacing, and interleaving for different contexts.

The Fringe

Ideas outside the mainstream—worth understanding, even if not endorsing.

Radical unschooling

Self-directed education advocates argue children naturally optimize their own learning when given freedom. Research shows unschooling can support psychological needs for autonomy and competence.

• The appeal: If adaptability includes meta-learning—knowing how you learn best—shouldn’t children discover this through experimentation?
• The critique: Self-directed doesn’t mean self-optimizing. Children may avoid domains requiring persistent effort through discomfort—exactly the desirable difficulties that build robust learning.
• Worth considering: Even within traditional schooling, some domains of genuine self-direction may build metacognitive capacity.

The anti-strategy position

Some argue explicit strategy instruction overcomplicates a natural process. Children have learned for millennia without metacognition workshops.

• The evidence: Children in informal environments often develop sophisticated implicit strategies. Much learning is procedural—you learn to ride a bike through practice, not metacognition.
• The tension: Implicit learning has limits. Complex academic domains and transfer to novel contexts may require explicit metacognitive engagement.
• Worth considering: Strategy instruction can become over-scaffolding. Sometimes the best intervention is challenge plus time.

AI as meta-learning partner

A provocative position: perhaps AI’s greatest educational value is teaching how to learn. AI that forces articulation of reasoning, identifies misconceptions, and models metacognitive questioning might build adaptability rather than undermine it.

• The evidence: Research on human-AI collaboration shows AI can enhance tutoring effectiveness. The right design might do the same for meta-learning.
• The risk: The same tools could easily smooth all friction and eliminate struggle. Design choices matter enormously.
• Worth considering: Rather than resisting AI in education, perhaps the focus should be demanding AI that builds capacity rather than creating dependency.

Resources

If you only do one thing after reading this:

Start here: Make It Stick: The Science of Successful Learning by Brown, Roediger, and McDaniel. The most accessible synthesis of cognitive science on effective learning—written by the researchers who did the work.

Contrarian pick: The Schools Our Children Deserve by Alfie Kohn. His argument: emphasizing outcomes undermines the intrinsic motivation that drives deep learning.

Books

• How We Learn by Benedict Carey — A journalist’s engaging tour through learning science, including counterintuitive findings about forgetting, sleep, and distraction.

• A Mind for Numbers by Barbara Oakley — Applies learning science to math and science, domains where many children struggle.

• Peak by Anders Ericsson — The science of deliberate practice from the researcher who defined it.

Research

• Bjork on Desirable Difficulties — The foundational paper. Conditions that slow initial learning enhance long-term retention.

• Roediger & Karpicke (2006) on the Testing Effect — Testing isn’t just assessment—it’s a learning event, often more powerful than study.

• 2024 Meta-analysis of Metacognitive Interventions — The latest synthesis: metacognitive interventions work (g = 0.48), especially when delivered by teachers.

Tools & Products

• Anki — Spaced repetition software implementing evidence-based learning algorithms. Free.

• Learning How to Learn (Coursera) — The most popular MOOC ever, with a youth version. Translates research into strategies.

• Quizlet — Flashcard platform with built-in spacing and retrieval practice.

Researchers to follow

• Robert Bjork — Desirable difficulties. UCLA.
• Barbara Oakley — Learning science for popular audiences.
• Daniel Willingham — Cognitive science applied to education.

Field Notes

Personal reflections and experiments coming soon. Subscribe to get notified when they’re published.

Last updated: 2025-11-29 Status: 🌳 Mature