Agency

TLDR: Agency is the belief that you can just do things—that you can act on the world without waiting for permission. In an AI age, it’s the difference between children who initiate and those who wait to be prompted.

Related Doing capacities: Persistence, Adaptability

Here’s something that should make every well-meaning parent uncomfortable: the more you help your child, the less capable they may become.

Researchers have found that children whose parents consistently step in to solve problems—tying shoes, resolving conflicts, completing difficult homework—show measurable decreases in their belief that they can handle challenges independently. The helping creates a dependency loop. Each intervention sends an implicit message: You can’t do this without me.

The phenomenon is called learned helplessness. Martin Seligman first documented it in the 1960s, studying what happens when organisms learn that their actions don’t affect outcomes. They stop trying. But the uncomfortable finding for parents: you can create learned helplessness through too much help, not just through adversity.

Seligman didn’t set out to study agency. He stumbled into its opposite. Running experiments on dogs and electric shocks at Penn, he noticed something that changed his career: dogs who learned they couldn’t escape shocks eventually stopped trying, even when escape became possible. He realized the same phenomenon explained depression in humans.

Then he reversed course. If helplessness could be learned, could agency be learned too? His pivot from studying what makes people helpless to what makes them thrive launched the positive psychology movement. The man who discovered how to break agency spent the rest of his career figuring out how to build it.

This paradox sits at the heart of raising children. Parents want kids who believe they can just do things. Kids who see themselves as capable actors rather than passive recipients of whatever happens to them. Yet protective instincts, amplified by an anxious culture, often lead to exactly what undermines this belief.

The research reveals a deeper tension: the capacity to initiate action doesn’t emerge from being told you’re capable, or even from succeeding. It emerges from the experience of struggling through challenges and discovering—through direct experience—that you can influence outcomes.

You cannot give a child agency. You can only avoid taking it away.

What Agency Actually Is

Agency is the belief that you can just do things.

Think of it this way: some kids see a problem and think “I could try something.” Other kids see the same problem and wait for someone to tell them what to do. That gap—the willingness to act without waiting for permission—is agency.

It breaks down into four pieces:

• Autonomy — The freedom to make choices. Not just having options, but experiencing those choices as genuinely yours rather than coerced or controlled.

• Self-efficacy — The belief that you can act effectively. Bandura’s research shows this is domain-specific (high self-efficacy for math, low for public speaking) but contributes to a general sense of capability.

• Intentionality — Acting with purpose rather than merely reacting. The difference between choosing to do something and being swept along by circumstances or others’ agendas.

• Ownership — Taking responsibility for outcomes. Seeing results as consequences of your choices, not things that happen to you.

Common confusions

People often mix up agency with related ideas. Here’s the difference:

“My kid has a good attitude—isn’t that agency?” Not quite. Believing your actions matter (locus of control) is related but narrower. Agency is the capacity to act, not just the belief that acting matters.

“Isn’t this just self-esteem?” No. A child can feel good about themselves but still wait for permission to start things. Agency is specifically about initiation—the willingness to act without being prompted.

“What about grit? Perseverance?” Different capacity. Agency gets you off the starting line. Persistence keeps you going when it gets hard. A child might easily start projects (strong agency) but abandon them when they hit resistance (weak persistence). Or vice versa—reluctant to start but tenacious once committed.

Here’s how I think about it: Agency gets you off the starting line. Persistence keeps you going when it’s hard. Adaptability helps you change course when needed. Three capacities, one flow: start → continue → adjust.

The key insight

Agency isn’t something you build into a child from scratch. Children are born with it. An inherent drive to affect their world. The “terrible twos” aren’t a problem to be managed; they’re agency asserting itself.

The question isn’t how to create agency. It’s how to avoid crushing it.

And there’s a new threat making that harder than ever.

The AI Complication

Every technology reshapes what it means to be an effective human. Writing externalized memory. Calculators externalized arithmetic. The question with AI is: what gets externalized, and what happens to the parts of us that relied on doing those things ourselves?

The threat to agency is subtle but structural.

AI reduces friction

AI can write your email, plan your trip, solve your problem. Each friction-reduction is a micro-decision you didn’t have to make, a challenge you didn’t have to navigate. For adults with established agency, this is efficiency. For children still developing the belief that they can do hard things, it’s a problem.

Agency is built through encountering resistance and pushing through it. Frictionless environments don’t build agency. They atrophy it.

AI creates dependency loops

A 2024 MIT Media Lab study found that students using ChatGPT exhibited the lowest brain engagement of any group studied, often copying AI output directly without cognitive processing. The deeper issue is what repeated use does to a student’s sense of themselves as a capable thinker.

Each time you outsource thinking, you strengthen the neural pathway that says I need help with this.

AI shapes choice invisibly

Recommendation algorithms decide what children see, effectively narrowing their world to a personalized bubble. The child feels like they’re choosing, but the choice set has been curated. This is a direct attack on agency: the subjective experience of choice without the reality of it.

Children raised in algorithmically curated environments may develop “pseudo-autonomy.” They feel free but aren’t.

AI substitutes for struggle

The marshmallow test for AI is homework. When a child can get an AI to do their math problems, write their essays, or solve their coding challenges, what’s the incentive to struggle through?

The struggle is the point. That’s where agency develops. AI offers an easy exit at exactly the moment when persistence would build capacity.

The neuroscience of AI-induced agency loss

The threat isn’t just behavioral—it’s neurological.

The comparator model problem. The sense of agency emerges when the brain’s prediction of an action’s outcome matches the actual sensory feedback. This requires you to initiate the action. When AI acts on your behalf—writing your paragraph, solving your equation—there’s no efference copy, no prediction, no match. The action-outcome loop that builds agency never fires. You experience the outcome without the neurological signature of having caused it.

The dopamine reinforcement gap. The dopaminergic system learns through action-outcome contingency: I did X, Y happened, that felt good, I’ll do X again. When AI provides the outcome without requiring the action, the reinforcement learning that builds agentic habits doesn’t occur. Worse, you may get the dopamine hit from the outcome (the completed essay, the solved problem) while the neural pathway connecting your effort to that outcome remains unstrengthened.

The prefrontal cortex development window. The PFC—seat of planning, decision-making, and impulse control—develops through use. Children who outsource cognitive work to AI during critical developmental windows may be literally shaping their brains toward dependency. The concerning parallel: children who don’t exercise visual processing during critical periods never develop normal vision. What happens to executive function if it’s never exercised?

Intentional binding and AI tools. Patrick Haggard’s research shows that people perceive voluntary actions and their outcomes as temporally closer together—a signature of agency. When AI mediates the action, this binding may not occur. The child experiences outcome without ownership.

The paradox

AI also expands what’s possible. A child with an AI assistant can build things, explore ideas, and create at levels previously impossible. The ceiling goes up. The question is whether the floor also goes up—or whether some children use AI to amplify their agency while others use it to replace their agency entirely.

The determining factor will likely be whether children have developed strong agency before AI becomes a constant presence. A child who already believes they can do hard things will use AI as a tool. A child who never developed that belief will use AI as a crutch.

Screen time beyond AI

AI is the sharpest threat to agency, but not the only one. The broader screen ecosystem affects agency through distinct mechanisms.

Passive vs. active consumption matters. Research on screen media types shows that passive screen use—watching videos, scrolling feeds—correlates with lower autonomy scores in children, while interactive use has more complex effects. The distinction isn’t screen vs. no-screen. It’s whether the child is doing something or receiving something.

A child building in Minecraft exercises agency. A child watching Minecraft videos does not.

Social media creates pseudo-agency in adolescents. A systematic review applying SDT found that social media both supports and thwarts adolescents’ autonomy needs. Platforms offer self-expression and choice, but simultaneously impose social pressures that create obligation and reduce genuine agency. The adolescent feels in control of their feed while algorithmic curation shapes their reality. Studies on passive scrolling link this behavior to increased psychological distress—scrolling is action without outcome, movement without mastery.

Video games present a genuine paradox. SDT-grounded research shows well-designed games can satisfy autonomy, competence, and relatedness needs. Players make meaningful choices, develop skills, connect with others. Games that support these needs are associated with positive well-being.

But research also shows that when adolescents’ real-world needs for autonomy and competence go unmet, they may turn to gaming as compensation—getting their agency fix virtually rather than developing real-world agentic capacity. The question isn’t whether games build agency, but whether they substitute for or supplement agency development in other domains.

The practical implication: Not all screen time is equal. Parents should distinguish between:

• Active creation (coding, building, designing)
• Interactive challenge (well-designed games with genuine choices)
• Passive consumption (watching, scrolling)
• Algorithmic immersion (recommendation-driven content)

The first two can support agency. The latter two typically undermine it.

The Research: What We Know

This isn’t one study or one researcher’s theory. The patterns replicate across dozens of studies and thousands of kids.

When parents give kids real choices, those kids do better. They develop better self-regulation, function better socially, and perform better academically. The effect sizes are medium but consistent (r = 0.33 to 0.46). This replicates.

Helicopter parenting backfires. Fifty-two studies later, the pattern is clear: overparenting is associated with more depression, more anxiety, and lower life satisfaction in kids. The cruelest irony? A 2024 meta-analysis found helicopter parenting associated with reduced self-efficacy (r = -0.21)—the very capacity parents think they’re protecting.

Agency can be developed. This is the hopeful part. Interventions that teach self-regulation actually work. Effect sizes of d = 0.42 to 0.64—in education research, that’s meaningful. Agency isn’t fixed at birth.

Kids who believe they control outcomes do better in school. Meta-analyses consistently find this (d = 0.36 to 0.45). The belief that your actions matter becomes self-fulfilling.

Self-control in childhood predicts adult life outcomes. The Dunedin study followed 1,000 people from birth to age 45. Children with lower self-control were 2.5 times more likely to have multiple health problems by their 30s. This held even after controlling for intelligence and socioeconomic status. Early agency casts a long shadow.

Rewards can undermine the thing they’re trying to build. 128 studies found tangible rewards decreased intrinsic motivation (d = -0.36).

Surprising Finding: Rewards Hurt Children More Than Adults

The undermining effect was stronger in children than college students. Children are particularly vulnerable to having their intrinsic motivation hijacked by external rewards—sticker charts, payment for grades, prizes for reading. The well-intentioned reward system may be systematically eroding the very agency it’s trying to build.

Executive function responds to training. Studies show preschool interventions can improve working memory, impulse control, and mental flexibility—all building blocks of agency. Effect sizes are medium (d = 0.30 to 0.47), which in education research is actually pretty good.

Physical activity builds agency. A 2023 meta-analysis found moderate-intensity exercise improved working memory (large effect), cognitive flexibility (large effect), and inhibitory control (medium effect). The body builds the brain’s capacity for agency.

For deeper treatment of mechanisms, debates, and frontier research, see the “Research: Going Deeper” section at the end of this document.

Early Childhood (0-5)

What we know

Agency emerges earlier than most people realize. By 2-3 months, your infant already recognizes that their actions cause effects. Kicking a leg makes a mobile move. This isn’t conscious agency, but it’s the neurological foundation.

Surprising Finding: Agency Begins at 2-3 Months

EEG research shows infants as young as 2-3 months detect contingencies between their actions and environmental events. When a baby kicks and a mobile moves, their brain is already wiring the connection: I did that. The earliest parent-child interactions are already shaping agentic development.

The “terrible twos” represent agency’s first dramatic assertion. Erik Erikson identified this period (18 months to 3 years) as “Autonomy vs. Shame and Doubt.” The child discovers they’re a separate being with a will of their own. The constant “No!” and insistence on “Me do it!” aren’t defiance. They’re agency emerging.

Parents who support this (within safety limits) lay the foundation for autonomous functioning. Parents who crush it with constant control may create lasting doubt.

By ages 3-6, children seek initiative. They want to start activities, make plans, try adult activities. This is Erikson’s “Initiative vs. Guilt” stage. The key is providing opportunities for child-initiated activity while setting limits that don’t shame the impulse itself.

The critical balance: Research shows that letting children take appropriate risks—climbing, exploring, making mistakes—builds agency. Over-protecting deprives them of the experiences that teach them they can handle challenges.

What you can do

• The Choice Protocol. Provide two or three acceptable options rather than open-ended questions or no choice at all. “Red shirt or blue shirt?” supports agency. “What do you want to wear?” can overwhelm. Dressing them with no discussion undermines agency. Research shows even small choices increase engagement.

• The Struggle Window. When children face frustration, wait. Let them struggle before offering help, then offer the minimum help needed. Studies found parents who learned to wait had children with better self-regulation. The temptation to swoop in comes from adult discomfort. But the struggle is where agency forms.

• Mastery Over Praise. Bandura’s research identified mastery experiences—successfully completing challenging tasks—as the most powerful source of self-efficacy. This means providing appropriately challenging activities and letting children complete them. A puzzle they solve themselves builds more agency than ten puzzles solved together.

• The Risk Permission. Explicitly permit physical risk-taking within assessed bounds.

  Instead of: “Be careful! Don’t fall! Hold on tight!” Try: “You can climb that tree—I’ll be here if you need me.”

  The first communicates doubt. The second communicates trust.

• Minimum Effective Help. When you do help, provide the smallest intervention that allows them to continue. Don’t tie the whole shoe—loosen the knot so they can try again. Don’t solve the puzzle—point to a piece they haven’t tried. The goal is enablement, not completion.

Middle Childhood (6-11)

What we know

As your child enters school, agency expands beyond “Me do it!” to include beliefs about competence across domains. They’re comparing themselves to peers, receiving formal feedback, building (or losing) confidence in specific areas.

Decision-making autonomy increases gradually through this period. Kids start managing more of their own time, choices, and activities.

Here’s the danger: excessive structure. School schedules, homework, activities. Children can easily become executors of adult-designed agendas rather than agents of their own lives. The antidote? Make sure they have some domain where they direct their own effort, make their own choices, and experience the results.

Domain-specificity emerges here. A child might feel highly agentic in art but helpless in math. These beliefs tend to persist. Early experiences of mastery or helplessness in a domain shape long-term orientation toward it.

What you can do

• The Consultant Stance. This is the single most useful reframe from Stixrud & Johnson: stop being your kid’s manager, start being their consultant. Managers direct. Consultants advise when asked.

  Instead of: “Did you finish your homework?” Try: “What’s your plan for getting your homework done?”

  The difference sounds small. It’s not. One puts you in charge. The other puts them in charge—with you available as a resource.

• Real Responsibility. Give children genuine responsibilities with real consequences—managing a budget for their activities, planning a family meal, caring for a pet. The key word is “genuine.” The responsibility has to matter, and failure has to be possible. Pseudo-responsibilities (“You’re in charge of watering the plant, but I’ll remind you”) don’t build agency because there’s no real ownership.

• The Rationale Practice. When setting limits, explain why. Research shows rationales help children internalize requests rather than merely comply.

  Instead of: “Because I said so.” Try: “I need you home by dinner because we’re eating together as a family.”

  Both set the same limit. One invites compliance. The other invites understanding.

• The Domain Handoff. Identify one area of life and transfer full ownership. Bedroom organization, extracurricular schedule, morning routine. They make the decisions, live with the consequences, learn from mistakes—without management. The domain should be meaningful enough to matter but bounded enough to be safe.

• The Consequence Preview. Before decisions, help them think through outcomes without dictating the choice. “If you spend all your allowance today, what happens when your friend’s birthday is next week?” Then let them decide—and live with it.

Adolescence (12+)

What we know

Adolescence brings a neurological shift that explains a lot: the reward system matures faster than the cognitive control system. This creates heightened risk-taking. But here’s the reframe—risk-taking is fundamentally an expression of agency. Your teenager is asserting that they can act on the world, even when those actions are unwise.

The developmental task of this period is integrating agency with identity. Who am I? becomes entangled with What can I do? and What do I choose to do? Autonomy-seeking intensifies, but research shows this doesn’t mean detachment from you. Adolescents still seek adult guidance. They want autonomy with connection.

Positive risk-taking—trying new activities, pursuing challenging goals, forming new relationships—is associated with healthy development. The goal isn’t to eliminate risk-taking but to channel it toward growth. Teens who lack opportunities for positive risk often seek it through negative means.

What you can do

• Graduated Autonomy. Identify domains where the adolescent can take increasing control—curfew negotiation, academic decisions, spending money. Start with lower-stakes domains and expand as they demonstrate capability. The goal: by 18, they’re managing most aspects of their own life.

• Process Over Outcome. When things go wrong, focus on process. “Walk me through your thinking” rather than “Here’s what you should have done.” This maintains their sense of agency even in failure. They made a decision, they can learn from it, they’ll decide differently next time.

• Failure Acceptance. Allow natural consequences to occur. Poor study habits lead to poor grades. Mismanaged money means they can’t buy what they want. The temptation to protect from consequences undermines the feedback loop that teaches agency. The job is helping them process consequences, not prevent them.

• The Decision Debrief. After major decisions (good or bad outcomes), have a no-judgment conversation about the process. “What information did you have? What did you consider? What would you do differently?” This builds metacognition without second-guessing their agency.

• The Safety Net Principle. Be a safety net, not a harness. A harness prevents falls. A safety net catches you after you’ve fallen. Make clear you’ll be there if things go seriously wrong, but won’t prevent the ordinary failures that build capability. The security of knowing you’re there enables risk-taking.

At Any Age

Some practices work whether your kid is 3 or 13.

• Model Agency (The Visible Struggle). Children learn by watching. When facing challenges, narrate your thinking: “This is hard. I’m going to try this approach.” Let them see you struggle and persist. Visible agency shapes their implicit beliefs about what’s possible.

• Acknowledge Perspective. Before requests or limits, acknowledge their viewpoint. “I know you want to keep playing, and it’s time to leave.” This simple practice—validated by extensive SDT research—maintains relationship while setting limits.

• Watch Your Language. Controlling language (“You should,” “You have to,” “You must”) undermines autonomy. Autonomy-supportive language (“You might consider,” “What do you think about,” “One option is”) maintains it.

Surprising Finding: The Words You Use Literally Change Outcomes

Research by Vansteenkiste and colleagues shows that when parents use autonomy-supportive language (explaining rationales, acknowledging feelings, offering choice), children show higher engagement, better learning outcomes, and more persistence. The same request, framed differently, produces different results. “You have to clean your room” vs. “I’d appreciate it if you could clean your room before dinner—it helps us all relax in a tidy space” aren’t just stylistically different. They produce neurologically and behaviorally different responses.

• The Control Audit. Periodically review where you’re controlling vs. supporting. Ask: “Where am I making decisions my child could make? Where am I solving problems they could solve? Where am I preventing failures they could learn from?” This self-assessment prevents control creep—the gradual expansion of parental management that happens without intention.

Special Considerations

Neurodivergent children

The autonomy-support research was largely conducted with neurotypical populations. The principles hold for neurodivergent children, but implementation requires adaptation.

ADHD: Autonomy support works, with scaffolding. Research shows children with ADHD symptoms who receive high parental autonomy support show better task perseverance than those who don’t.

Surprising Finding: ADHD Children May Benefit Most from Autonomy Support

Counterintuitively, children with ADHD symptoms who receive high parental autonomy support show better task perseverance than neurotypical children without such support. The expectation might be that ADHD requires more control, more structure, more external management. The evidence suggests the opposite: autonomy support may be especially powerful for children whose executive function deficits make external control feel particularly aversive.

The key is that autonomy support must coexist with external structure that compensates for executive function deficits. Interventions like Cog-Fun combine autonomy support with explicit executive function training. Rather than tightening control for ADHD children, parents might get better results by providing structure and autonomy—clear scaffolding combined with genuine choice within that structure.

Autism: Self-determination is achievable. Research shows teacher autonomy support correlates with higher self-determination in autistic children. Self-monitoring interventions—teaching children to observe and record their own behaviors—effectively build independence.

Autistic children can develop strong agency but may need more explicit instruction in component skills that neurotypical children acquire implicitly.

Anxiety: Calibration required. Experimental studies show parental control increases anxious children’s negative predictions and distress. Yet other research finds anxious children with low perceived control who receive high autonomy may show increased physiological stress.

A University of Washington study found parenting styled to temperament cut anxiety and depression in half: children with good self-regulation thrived with high autonomy, while children with lower self-regulation needed more structure.

The meta-principle: Neurodivergent children need autonomy support as much as—perhaps more than—neurotypical children. But “support” may require more explicit scaffolding, more environmental structure, more calibration to the individual child. The goal remains the same. The path requires adaptation.

Gender

Studies across cultures find parents often grant more autonomy to girls than boys—potentially because girls are perceived as more self-regulated, or because boys’ more disruptive behaviors prompt more controlling responses.

Research on parent-adolescent perceptions shows boys may be more sensitive to parental negativity and controlling behaviors. Boys’ depression appears more strongly predicted by autonomy support discrepancies.

In families with traditional gender attitudes, research shows girls receive fewer autonomy opportunities—agency development constrained by gendered expectations.

The implication: Examine whether you’re providing equivalent autonomy support across genders. Don’t assume boys need more control or girls need less challenge.

Where Things Go Wrong

The help trap

You see your kid struggling. Every instinct says help. Here’s the problem: that instinct, followed too often, creates exactly what you’re trying to prevent.

The research is clear: parental over-involvement—helicopter parenting, snowplow parenting, lawnmower parenting—predicts worse outcomes. Children of over-involved parents show higher anxiety, higher depression, lower life satisfaction, poorer coping skills, diminished self-efficacy.

The mechanism is straightforward: by removing obstacles, you remove the experiences through which agency develops. Each “rescue” sends a message: You can’t handle this. Repeat it enough, and your child believes it.

The insidious part is that this comes from love. You help because you don’t want your child to suffer. But appropriate suffering—struggle, frustration, failure—is the raw material of agency. You cannot build the capacity to handle difficulty without handling difficulty.

The opposite mistake: no structure at all

Some parents hear “autonomy support” and think it means “let them do whatever they want.” This creates different problems.

Children need structure—clear expectations, consistent limits, predictable routines. Structure provides safety and, paradoxically, enables autonomy. When limits are clear, kids can operate freely within them.

Autonomy-supportive parents have high expectations and clear limits. What makes them autonomy-supportive is how they enforce those expectations: through explanation, choice, and acknowledgment of the child’s perspective—not coercion, control, and punishment.

Agency without judgment

There’s a shadow side. Agency without Judgment is recklessness. A kid who believes they can do anything but can’t evaluate risks? That’s dangerous—to themselves and others.

The adolescent taking up extreme sports without safety knowledge. The child challenging authority destructively. The teen making impulsive decisions with long-term consequences. Agency needs its siblings: Judgment to evaluate, Persistence to follow through.

Structural barriers

Agency development can be undermined by systems beyond parental control:

• Schools that offer little choice, emphasize compliance, and evaluate narrowly can suppress agency even when parents support it.
• Over-scheduling leaves no time for child-directed activity—the primary venue for agency development.
• Technology ecosystems designed to maximize engagement create dependency that undermines self-directed action.

Parents must work within these constraints while advocating for change.

The Research: Going Deeper

Everything above is the practical summary. If you want the full scientific treatment—mechanisms, debates, frontier research—keep reading. If not, skip to the Resources section below.

The mechanism

Agency has a neurological signature. The sense of being the cause of your actions involves the comparator model, first proposed by neuroscientist Patrick Haggard.

When you initiate a voluntary action, your brain generates a motor command and simultaneously creates an “efference copy”—a prediction of what sensory feedback the action will produce. The actual feedback is compared to the prediction. When they match, you experience agency: I did this. When they don’t match, you experience that action as not your own.

This explains why agency requires action. You can’t feel like an agent by watching others act. The prediction-feedback loop requires you to initiate.

The prefrontal cortex—the brain region responsible for planning, decision-making, and impulse control—is central to agency. It develops slowly. Basic executive functions emerge in early childhood, but sophisticated self-control continues developing into the mid-twenties.

Children can’t exercise adult-level agency because the neural hardware isn’t there. But this doesn’t mean agency can’t be developed. It means it must be scaffolded appropriately—providing external structure the child’s brain can’t yet provide, then gradually transferring control as capacity develops.

The dopaminergic system also matters. Dopamine isn’t just about pleasure—it’s about motivation and anticipation of reward. When children experience success from their own efforts, dopamine reinforces the connection between action and outcome. When success is unearned, the reinforcement doesn’t happen. This is why praising effort over outcomes matters: reinforcing the action component of agency, not just the result.

Where experts disagree

The Marshmallow Test Controversy. Walter Mischel’s iconic study has been substantially challenged. In 2018, Tyler Watts and colleagues replicated with a larger, more diverse sample (918 vs. 90 children). The predictive power largely disappears when controlling for socioeconomic status, family background, and early cognitive ability.

What looks like individual willpower may be substantially determined by environment. A child from a stable, resource-rich environment has reason to trust that the second marshmallow will appear. A child from an unpredictable environment has learned that waiting often means losing. The “failure” to delay gratification may be rational adaptation, not deficit.

The Free Will Problem. Benjamin Libet’s experiments found brain activity for voluntary movement begins ~550 milliseconds before people report consciously deciding to move. Conscious awareness comes ~200ms before movement—350ms after the brain has initiated.

If brains start actions before we “decide,” what does that mean for agency? Libet himself proposed we retain “veto power.” The practical implication: agency may not be about conscious control of every action, but about building the habits and values that shape what our brains initiate.

The Structural Critique. Critical theorists argue the Western emphasis on individual agency is culturally specific and potentially harmful. By framing outcomes as results of individual choice, agency narratives obscure systemic barriers and blame individuals for structural failures.

A child in poverty, attending under-resourced schools, facing discrimination—does telling them to “take agency” help or harm? SDT researchers respond that autonomy support works across socioeconomic contexts. But the critique has teeth. Be wary of agency-talk that ignores real constraints.

Too Much Autonomy Support? Research by McCurdy (2022) found higher maternal autonomy support associated with increased anxiety in children.

Surprising Finding: Too Much Autonomy Support Can Backfire

This complicates the “more autonomy is always better” narrative. Verhoeven et al. (2012) found parental psychological autonomy support linked with increased child anxiety in children aged 10 or younger. The relationship may be curvilinear: both too little and too much autonomy support can be harmful. Children need appropriate scaffolding, not abandonment to figure everything out alone.

Collective vs. Individual Agency. The Western framing assumes the individual is the relevant unit. But research on relational and collective agency challenges this. Children’s actions are always embedded in relationships. Agency may be better understood as something that exists between people, not just within individuals.

Studies of Swedish children found they perceived the most agency in peer interactions and the least with teachers—suggesting agency is highly context-dependent and relational. Cross-cultural research shows that in collectivist societies, agency often means acting through and for relationships, not independently of them.

The Replication Question. How robust is this research? The meta-analyses are encouraging, but individual studies vary in quality. The field hasn’t had the systematic replication efforts that challenged other areas of psychology. Effect sizes are modest (r = 0.33-0.46 for parenting autonomy support on grades). This doesn’t mean the research is wrong—but intellectual honesty requires noting that some findings may not replicate as cleanly as summaries suggest.

The frontier

Intentional binding. Haggard’s work opens new ways to measure agency objectively. When you perform a voluntary action producing an outcome, you perceive action and outcome as temporally closer together. This “intentional binding” is specific to voluntary actions. Researchers now use it as an implicit measure, potentially allowing study of agency development in infants who can’t self-report.

Embodied agency. Emerging research connects interoception—perceiving internal bodily signals—to agency development. Studies show interoceptive sensitivity is present from infancy and linked to self-awareness. Agency may be fundamentally embodied.

This connects to physical activity and risky play.

Surprising Finding: Risky Play Builds More Than Physical Skills

Systematic reviews reveal something counterintuitive: children who climb trees, jump from heights, and explore unfamiliar terrain don’t just develop physical skills—they show higher creativity scores, better social skills, and greater resilience. The mechanism: physical mastery provides immediate, undeniable feedback that their actions matter. The body knows it climbed the tree.

Research on risk willingness shows children in physically challenging play develop better risk assessment and movement adaptability. Studies comparing play environments find children in challenging, unpredictable settings acquire better motor control than those in structured environments. The irony: safety measures we impose (padded playgrounds, constant supervision, prohibiting climbing) may undermine the very psychological resilience we’re trying to protect.

AI as agency amplifier. Could AI build agency if designed differently? Imagine AI that requires children to make decisions, scaffolds struggle rather than eliminates it, makes the child’s agency visible. The Generation AI project in Finland explores how teaching children about AI can enhance autonomy.

Computational models. Researchers are building models simulating how the sense of agency emerges through action-feedback loops. These predictive processing frameworks may eventually inform the design of educational technology and AI systems.

The Fringe

Good research means engaging with positions that push beyond mainstream consensus. These ideas are provocative—worth understanding, even if you don’t endorse them.

Taking Children Seriously (TCS)

This philosophy argues children should have full autonomy over all decisions—screen time, food, sleep, education. Founded by Sarah Fitz-Claridge and influenced by Karl Popper, TCS holds that coercion is always harmful.

• The appeal: Takes autonomy research to its logical extreme.

• The critique: Ignores developmental reality. Children’s prefrontal cortices aren’t developed for adult-level decision-making. Research on structure suggests children need limits. Most developmental psychologists would consider TCS an overcorrection confusing autonomy support with abdication.

• Worth considering: TCS’s emphasis on taking children’s preferences seriously and avoiding casual coercion has value, even if the full philosophy goes too far.

The case against homework

Peter Gray and others argue homework fundamentally undermines agency by extending school control into home life and reducing time for self-directed activities.

• The evidence: Research on homework effectiveness is surprisingly mixed. For elementary students, homework shows minimal academic benefit. The costs—family stress, reduced play time, potential agency damage—may outweigh gains.

• Worth considering: Even without eliminating homework, recognizing its agency costs and advocating for less/better homework is defensible.

Radical unschooling

The extreme version of self-directed education: children control all aspects of their lives and learning, with no required curriculum.

• The evidence: Surveys of unschooled adults show many pursue higher education and feel well-prepared. However, other research finds unschooled children may struggle with structured content and subjects requiring persistent effort.

• The tension: Radical unschooling may build agency at the cost of competence in domains the child doesn’t spontaneously choose.

Resources

If you only do one thing after reading this:

Start here: The Self-Driven Child by Stixrud & Johnson. The most accessible synthesis with practical applications. The “consultant not manager” reframe alone is worth the read.

Contrarian pick: Visit an adventure playground or Sudbury school. These approaches seem radical but have theoretical grounding. Even if you don’t enroll your child, visiting can shift your sense of what kids are actually capable of.

Educational approaches

• Montessori — Children choose activities within a prepared environment. A 2025 RCT showed Montessori preschoolers outperformed peers in executive function, reading, and social understanding.
• Reggio Emilia — Project-based, child-interest-driven; teachers as facilitators rather than directors.

Products

• Open-ended toys — Blocks, LEGO, loose parts. No “right” way to play means kids must decide what to create.
• Maker tools — Arduino, Raspberry Pi. Building real things creates powerful mastery experiences.
• Creation software — Scratch, Tinkercad. Low-floor, high-ceiling tools that let kids create without external validation.

Programs

• Adventure playgrounds — Supervised spaces for managed physical risk-taking.
• Sudbury schools, Agile Learning Centers — Maximum autonomy in learning choices.

Books

• The Self-Driven Child (Stixrud & Johnson) — The accessible synthesis. Start here.
• Free to Learn (Peter Gray) — Provocative case for self-directed education.
• Hunt, Gather, Parent (Doucleff) — Cross-cultural perspective on raising autonomous kids.

Researchers to follow

• Deci & Ryan — Self-Determination Theory founders.
• Wendy Grolnick — Leading researcher on parental autonomy support.
• Peter Gray — Free play advocacy and the Freedom to Learn blog.

Field Notes

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

Further Reading

Foundational

• Deci & Ryan (2000). “The ‘What’ and ‘Why’ of Goal Pursuits.” — The manifesto on autonomy, competence, and connection.

• Bandura (2006). “Toward a Psychology of Human Agency.” — Fifty years of self-efficacy research synthesized.

• Grolnick (2009). “The Role of Parents in Facilitating Autonomous Self-Regulation.” — SDT theory to parenting practice.

Recent

• Haggard (2017). “Sense of Agency in the Human Brain.” — Definitive neuroscience review.

• Moffitt et al. (2011). “A Gradient of Childhood Self-Control Predicts Health, Wealth, and Public Safety.” — The Dunedin 45-year study.

• Watts et al. (2018). “Revisiting the Marshmallow Test.” — The replication that complicated the iconic finding.

Contrarian

• Markus & Kitayama (1991). “Culture and the Self.” — Is “agency” culturally constructed?

• Gray (2013). Free to Learn. — The radical case for self-direction.

• Kohn (1999). Punished by Rewards. — Against all external rewards.

Last updated: 2025-11-29 Status: 🌳 Mature Word count: ~8,900 Tags: #new-literacy #agency #autonomy #self-determination #parenting #capacities-framework