What Children Actually Need From Safety

The public conversation about children's safety, especially around digital life, is overwhelmingly a conversation about what to remove. Block this app. Restrict that platform. Raise the age at which a child gets a phone. Impose stricter content filters. Require age verification.

These are reasonable instruments at a population level, and I am not against them. But they share an assumption I think is mistaken: that safety is fundamentally a question of removal — taking dangerous things out of children's environments — and that once enough has been removed, what remains is, by default, safe.

Developmental psychology has a different picture. Safety, in the sense that matters for a young person's growth, is not the absence of risk. It is the presence of certain conditions, sustained over time, without which the rest of development cannot happen well.

What those conditions are

The shortest list I would defend is four things.

Predictable adults. A child does not need a perfect parent or teacher. They need an adult whose responses to them are predictable enough to model. The single most consequential variable in early regulatory development is whether the people around the child respond in roughly the same way today as they did yesterday — not necessarily warmly, not necessarily kindly, but coherently. This is the substrate on which everything else builds.

Repair after rupture. Adults will misread, snap, withdraw, fail. The question is not whether ruptures occur — they always do — but whether they are repaired. Children who watch ruptures get repaired learn that conflict is survivable. Children who watch ruptures stay open learn that they are not. The capacity for repair is a more useful safety variable than the absence of conflict.

A regulatory partner. Young children cannot self-regulate; they co-regulate. They borrow the nervous system of the adult next to them until their own is built. This is not a metaphor. It is a measurable physiological fact in the data on heart-rate variability, cortisol, and behavioural state. Older adolescents are still doing some version of this, just at a slower frequency and with peers as well as adults.

Permission to feel large feelings without losing the relationship. A child can be angry, frightened, or overwhelmed without that being treated as a fault to be punished. This is the variable that, in my reading, the moralised conversation about behaviour gets most wrong. Children whose intensity is read as bad character learn to suppress signal, not produce calm.

Why restriction-only safety fails

Restriction is necessary at the margins. A child should not be on a platform optimised for adult attention. A teenager should not be exposed to content that adults could not metabolise. Some doors should be closed, full stop.

But restriction, on its own, does not provide any of the four conditions above. A child who is removed from a dangerous platform but returned to a household whose adults are unpredictable is not, in the developmental sense, safer. They are just dangerous in a different room. And worse: the conversation about their safety has been resolved publicly, which can make the actual ongoing absence harder to see.

This is the part of children's safety politics that I think is most often missed. Headlines about restriction generate the appearance of having done the work. The work — the slow, expensive, unglamorous work of being a predictable adult who repairs ruptures and tolerates large feelings — is what actually produces safe children, and it is the part that policy cannot do for you.

What this means for digital systems

Digital and algorithmic environments are now part of the regulatory field, whether we acknowledge that or not. A teenager whose attention is held by a recommendation system is, for the duration of that holding, taking some of the system's qualities into their own state. A predictable, calmly-paced system contributes one thing to the child's regulatory load. A volatile, novelty-maximising system contributes another.

This is one of the questions my current research is examining. Early indications are that the systems most successful at maximising engagement are the systems that produce the most regulatory cost, and that this cost is heaviest precisely on the children with the least predictable adults at home — the children who could least afford it. The two failures compound.

Designing for safety means designing for repair, not for the impossibility of harm.

The implication

The reframe I would press on the conversation is this: we cannot keep children safe by subtracting things from their environments. We can only keep children safe by adding the things their development requires. The two strategies look similar from the outside; they produce very different children.

Restriction is the easier conversation to hold in public. The other one — the one about whether the adults in a child's life are present, attuned, repairable, and tolerant of large feelings — is the conversation that actually matters. We should be having it more.

Further reading

• Schore, A. N. (2003). Affect Regulation and the Repair of the Self. New York: Norton.
• Tronick, E. (2007). The Neurobehavioral and Social-Emotional Development of Infants and Children. New York: Norton.
• Hari, J. (2022). Stolen Focus. London: Bloomsbury.
• Twenge, J. M. (2023). Generations. New York: Atria.