By Fabrizio Rinaldi

The problem with optimising childhood is not that it makes children work harder. It is that it can leave them with fewer ways to be human when the plan stops working.

A few years ago, a parent asking me how to prepare a child for the future would usually receive an answer packed with useful nouns: coding, English, problem solving, STEM, perhaps a respectable extracurricular activity arranged with military precision between swimming and piano.

Today, I would start somewhere else.

I would ask what happens when the Wi-Fi fails, the plan changes, the group project falls apart, the teacher does not have the answer, or the AI gives a perfectly polished and confidently wrong response. I would ask whether the child can pause, look around, ask for help, invent an alternative and still trust their own mind.

This is not an argument against competence. It is an argument against confusing performance in stable conditions with the capacity to remain capable when conditions stop being stable. In a world designed around maximum efficiency, this distinction is becoming existential.

Olivier Hamant, a biologist and biophysicist, calls this the difference between performance and robustness. Performance is the ability to obtain the best result with the least possible resources. Robustness is the ability to remain viable through fluctuation, shock and uncertainty. In his reading, living systems do not primarily optimise for a single ideal condition. They survive by preserving options, diversity, redundancy and the capacity to adapt. (Source 1) (Source 2)

That insight should not remain inside an ecology lecture, a boardroom risk register or a LinkedIn post about supply chains. It should change the way we think about raising children.

---

The optimisation trap is not just an economic problem

We have spent decades treating friction as a design flaw. We made supply chains leaner, communications instantaneous, decisions more data-driven and childhood more scheduled. Every surplus minute became an opportunity to add a useful activity. Every pause became something that could be filled with content.

This has produced genuine benefits. Efficiency has helped us coordinate complex systems, reduce waste in specific processes and widen access to knowledge. The point is not to romanticise inefficiency, or pretend that a hospital should respond to an emergency with the relaxed rhythm of a Sunday lunch in Milan.

The problem begins when a tool becomes a worldview.

Hamant and Stéphane Grumbach describe suboptimality as a systems principle: locally inefficient processes can support resilience at the level of the whole. In their examples, the system gives up some immediate performance in order to preserve adaptability, variation and long-term survival. (Source 2) Hamant goes further in his recent work, arguing that robust systems often contain forms of internal tension, diversity and even apparent incoherence that prevent a single feedback loop from running unchecked. (Source 3)

This is counterintuitive because our institutions love clean dashboards. A child who finishes first, a school that raises test scores, a company that eliminates idle capacity, an AI assistant that removes every cognitive detour: these outcomes are all legible. They are also dangerously easy to mistake for health.

Logic of optimisation	Logic of robustness
Remove slack	Protect useful margins
Standardise one best path	Maintain multiple viable paths
Reward speed and predictability	Reward adaptation and recovery
Treat error as waste	Use error as information
Delegate complexity away	Learn to navigate complexity
Measure output	Build capacity

The key question is not whether something is efficient. The key question is: efficient for what, over what time horizon, and at whose expense?

This is where the old management adage known as Goodhart’s law becomes uncomfortable. When a measure becomes a target, it stops being a good measure. (Source 3) The child who learns to produce the correct answer fastest may be developing knowledge. They may also be learning that the purpose of thinking is to make uncertainty disappear as quickly as possible.

That is not the same thing as intelligence. It is certainly not the same thing as wisdom.

---

Childhood is not a KPI

We have built an adult world that treats bandwidth as a scarce resource. Meetings devour attention, notifications fracture it, and endless optimisation asks us to turn every spare minute into output. It is tempting to pass that logic down to our children because it looks like care.

We timetable them because we want to give them opportunities. We intervene because we want to protect them from frustration. We provide the answer because we do not want them to feel lost. We hand them a screen because we need ten quiet minutes to survive the late-afternoon chaos.

None of this makes parents negligent. It makes us human inside an economic system that has quietly removed the margins from family life.

But a child needs some margins.

The evidence from child development is not subtle on this point. Responsive, back-and-forth interaction with caring adults, what Harvard’s Center on the Developing Child calls serve and return, helps build the foundations for language, social skills and later cognitive development. (Source 4) Executive function and self-regulation, the capacities that help us plan, focus attention, switch strategies and hold information in mind, are not finished products delivered at birth. They are built over time, through developmental environments and repeated practice. (Source 5)

This has a very practical implication. Education for uncertainty cannot be reduced to teaching children to “be resilient” in the motivational-poster sense. It begins by giving them relationships stable enough to explore from.

A robust child is not a child who never needs help. It is a child who has experienced help without being robbed of agency.

The American Academy of Pediatrics frames play as a developmental necessity, not a reward after the serious work is done. Its clinical report, reaffirmed in 2025, connects developmentally appropriate play with social-emotional development, self-regulation and executive function. It also makes an argument that should worry every parent who has turned childhood into a logistics operation: play lets children test boundaries, negotiate rules, take proportionate risks and learn how to recover when outcomes are not fully controllable. (Source 6)

The future will not belong to children who never encountered friction. It will belong to children who can metabolise friction without interpreting it as personal failure.

---

The school of unnecessary things

There is a kind of learning that looks inefficient from the outside.

It is the child building a lopsided cardboard city with friends. It is a teenager repairing a bicycle badly before repairing it less badly. It is the student who spends an afternoon trying to solve a problem, gets it wrong, argues with classmates, and discovers that the wrong path has taught them what the right explanation needs to explain.

This is not wasted time. It is cognitive redundancy.

A society that has only one way of calculating, communicating, finding information or making decisions becomes fragile. The same is true of a mind. If a child has only one successful mode, memorise, comply, ask an adult, ask an app, they may become very effective within a narrow channel. But they will have fewer internal alternatives when that channel closes.

The educational research around productive failure is useful here, provided we do not turn it into another slogan. Manu Kapur’s model asks learners to struggle first with problems that exceed their current knowledge, then receive the instruction that lets them understand the gap between their attempted solution and a stronger one. The approach can improve learning transfer in suitable settings, but it is not a universal recipe and may not benefit young children or different learner profiles in the same way. (Source 7)

That caveat matters. Robustness is not the celebration of struggle for its own sake. Leaving a child alone in an impossible task is not education. It is abandonment wearing a productivity hoodie.

The adult’s job is to design safe difficulty. Enough challenge to generate curiosity and effort, enough support to prevent overwhelm, enough time for the child to discover that confusion is not an emergency.

This is close to the approach I call the Parent Orchestrator in my recent book, Crescere Umani. The point is not to keep children away from technology or to turn them into miniature engineers. It is to protect the conditions in which they become active, critical and creative participants in a digital world rather than passive recipients of its defaults. The book argues for physical play, curiosity, critical awareness of algorithms and AI as a sparring partner rather than a shortcut. (Source 11)

In other words: less “How can I remove every obstacle?” and more “Which obstacles are worth meeting, together?”

---

Teach children to keep more than one door open

Robustness is often described in terms of redundancy. In childhood, redundancy does not mean filling the week with seven different courses. It means giving a child more than one route into meaning and competence.

A child who draws, gardens, codes, argues, reads fiction, plays sport, cooks, looks after someone smaller and gets bored has more routes into the world than a child trained only to maximise a single score. They may not be the fastest in every domain. That is precisely the point.

The OECD Learning Compass 2030 frames education around student agency, well-being and the ability to find one’s direction responsibly in unfamiliar contexts. (Source 8) It is a useful image because a compass does not provide a pre-programmed itinerary. It gives orientation while the landscape changes.

Capability to cultivate	What it looks like at home or school	Why it adds robustness
Attention	Reading, making, conversation without a second screen	Protects the ability to notice before reacting
Emotional regulation	Naming frustration, pausing, returning to a task	Turns stress into a signal, not a command
Practical agency	Repairing, cooking, building, navigating offline	Creates alternatives when systems fail or change
Cooperation	Group projects, shared care, negotiated rules	Reduces dependence on the lone high performer
Systems thinking	Asking what a product, platform or choice affects elsewhere	Counters simplistic cause-and-effect stories
Critical judgement	Comparing sources, challenging an AI answer, stating uncertainty	Prevents fluent output from being mistaken for truth

These are not soft skills in the dismissive sense. They are the operating system underneath every technical skill that will matter five years from now.

They also require something our culture finds irritating: unstructured time. Boredom is not a bug in childhood. It is often the empty room in which curiosity finally has enough silence to speak.

---

AI should be the wind tunnel, not the autopilot

The arrival of generative AI makes this debate urgent because it can either widen or shrink a child’s repertoire of thought.

Used badly, AI becomes the ultimate optimisation device. It removes the first draft, the awkward question, the wrong answer, the blank page, the moment of checking whether a claim is actually true. It creates a seductive illusion of competence: beautifully formatted work, delivered at a speed that no child could sustain alone, with none of the intellectual muscle built along the way.

A child who outsources every difficult cognitive move may get better outputs while developing fewer capabilities. That is the educational version of a supply chain with one supplier, no warehouse and a CFO who has mistaken fragility for excellence.

Used well, however, AI can become a remarkable wind tunnel for thinking. A child can ask it for counterarguments, make it role-play a historical opponent, compare multiple explanations, identify a bias, generate several approaches to a problem, then test and improve them. The human task does not disappear. It becomes more demanding: evaluate, contextualise, verify, choose.

UNESCO’s AI Competency Framework for Students explicitly centres a human-focused mindset, ethics, technical understanding and system design, across the progression from understanding to applying and creating. It presents students as responsible citizens and co-creators, not merely consumers of a new interface. (Source 9)

This aligns with the wider educational direction outlined by the OECD. Agency is not independence from everyone. It is the capacity to act purposefully and responsibly in relation to others. (Source 8)

For parents and teachers, the practical test is simple. After using an AI tool, can the child explain what they accepted, what they rejected, what they changed and why? If not, the tool has probably optimised the task while hollowing out the learner.

That should also make leaders nervous. In my earlier piece on digital orchestrators, I argued that AI forces us to move from ambition to purpose. The same applies here. If an educational technology cannot answer what human purpose it protects, it is simply efficiency with a shiny user interface.

---

A robust society starts with the conditions around the child

It would be comforting to say that the solution lives entirely in the family home. Put down the phone, go to the park, have better conversations, buy fewer plastic objects that make noises no adult requested. All good ideas.

All insufficient.

A parent cannot build robust childhood alone while working in a culture that treats caregiving as an inefficiency, public spaces as a cost centre and teachers as content delivery systems. Harvard’s work on responsive relationships is clear that caregiver capacity is affected by financial pressure, health, social connection and the environment around the family. (Source 4) Robust childhood is therefore not a private lifestyle choice. It is a collective design problem.

UNESCO’s call for a new social contract for education makes a similar case at the institutional level. It frames education as a common good and places care, reciprocity, solidarity, social justice and respect for life at the centre of the project. (Source 10)

This is where the discussion has to move beyond “future-proofing” individuals. Nobody can be future-proofed. The phrase itself belongs to the fantasy that a child can be made invulnerable through the correct acquisition of credentials.

What we can build are future-capable communities.

A future-capable school does not simply add an AI module to an unchanged timetable. It creates time for inquiry, shared projects, repair, debate, outdoor exploration and reflection. It teaches statistical thinking but also asks what gets lost when every human question is turned into a metric. It treats technology as material for critique and creation, not as an automatic upgrade.

A future-capable city protects libraries, playgrounds, sports spaces, community workshops, green areas and public transport. These are not decorative amenities. They are the social equivalents of redundancy. They create places where people can learn, meet, repair, care and improvise outside the narrow transaction of buying and selling.

A future-capable company stops claiming that human development is a side benefit of employment. It creates room for learning, mentorship, rest and cross-functional thinking, because a workforce with no spare cognitive capacity is not efficient. It is one unexpected crisis away from a very expensive PowerPoint about resilience.

---

Five principles for raising a robust generation

The challenge is not to install a new perfect parenting system. That would be an impressively fast route back into the optimisation trap. The challenge is to use a few principles as a compass.

Principle	For parents and caregivers	For schools and institutions
Protect slack	Leave time for boredom, conversation and self-directed play	Design schedules with room for exploration, reflection and recovery
Stage safe difficulty	Let children attempt, fail and try again before intervening	Use guided inquiry and age-appropriate productive struggle
Expand repertoires	Mix digital, physical, social, creative and practical activities	Value diverse forms of competence, not only one score
Build reciprocal relationships	Practise attentive back-and-forth, not constant instruction	Treat belonging, trust and collaboration as learning infrastructure
Keep humans in the loop	Ask children to explain their reasoning after using AI	Assess judgement, process and revision, not only polished output

These principles do not promise control. They create something better: a larger range of possible responses.

That is the essence of robustness. Not invulnerability. Not heroic self-sufficiency. Not the endless performance of coping. A larger range of possible responses, held inside relationships strong enough to make experimentation safe.

---

The education we need is slower in the right places

There is a deep irony in all this. We keep trying to prepare children for an uncertain world by making their development more predictable. We measure earlier, schedule tighter, intervene faster and increasingly let machines complete the parts of thinking that feel slow, messy or emotionally inconvenient.

But uncertainty is not a technical glitch that will be eliminated by a better dashboard. It is the environment in which our children will live.

The alternative is not nostalgia. It is not anti-technology. It is not asking children to become little survivalists who can fix a toaster and identify edible plants while their classmates learn Python.

It is more demanding than that.

It means raising people who can live with a question before rushing to an answer. People who can use a machine without mistaking its fluency for wisdom. People who know that cooperation is not the consolation prize for those who failed to win alone. People who have been allowed enough play, enough attention, enough error and enough human contact to remain curious when the world refuses to behave like a spreadsheet.

Cui bono when a child becomes perfectly optimised for a future none of us can reliably describe?

And what might change if we stopped educating children to fit the machine, and started educating society to remain worthy of the humans growing inside it?

Sources

Source 1: The Great Simplification, Episode 230: The Optimization Trap

Source 2: Grumbach and Hamant, The case for suboptimal systems

Source 3: Hamant, Is incoherence required for sustainability?

Source 4: Harvard Center on the Developing Child, Serve and Return

Source 5: Harvard Center on the Developing Child, Executive Function

Source 6: American Academy of Pediatrics, The Power of Play

Source 7: Association for Psychological Science, Productive Failure

Source 8: OECD Learning Compass 2030

Source 9: UNESCO, AI competency framework for students

Source 10: UNESCO, Reimagining our futures together

Source 11: Fabrizio Rinaldi, Crescere Umani