It’s Not Just What a Platform Does. It’s who It does it to
Neurodivergence, digital design, and the children least protected by “choice”
It is too easy to say that children are addicted to screens.
It is easy because it gives us a familiar story: weak willpower, poor parenting, too much time online, not enough discipline, not enough fresh air. It gives schools something to say, parents something to feel guilty about, policymakers something to gesture towards, and platforms something to avoid.
But the word “addiction” can hide as much as it reveals. It can make the child look like the problem, rather than the environment they have been placed inside.
A child I know is autistic, ADHD and trauma-impacted. He is drawn to high-stimulation, reward-based content: drill rap, GTA videos, intense gaming, status-driven online worlds, fast-moving clips, conflict, danger, repetition, rhythm, sensation. He is also impulsive, rigid, loyal, easily overwhelmed, deeply sensitive to humiliation and often misunderstood. Offline, these traits already need careful support. Online, they meet systems designed to intensify them.
He is not simply “addicted to screens”. He is responding to an environment built to hold onto him.
And the question we need to ask is not only what platforms do. It is who they do it to.
A design feature that is annoying for one child may be destabilising for another. Autoplay, infinite scroll, notifications, streaks, loot boxes, algorithmic recommendations, personalised feeds, viral conflict, aggressive marketing, social comparison and reward loops do not land evenly across all users. They act on children with different nervous systems, different histories, different capacities to pause, different relationships to risk, different levels of loneliness, shame, impulsivity, sensory-seeking, anxiety or social vulnerability.
For many neurodivergent children and young people, digital spaces can be lifelines. They can offer friendship, identity, humour, special interests, creativity, music, gaming, belonging, information and relief from the exhaustion of face-to-face social performance. For autistic children, online life can sometimes reduce the pressure of eye contact, body language and sensory overload. For children with ADHD, fast-moving digital spaces can feel engaging in a world where school often feels like failure. For children who are isolated, bullied or excluded, the online world may be the place where they feel most competent.
That is why this argument cannot be reduced to “screens are bad”. The problem is not the existence of digital life. The problem is the commercial architecture of much of it.
When platforms are designed around attention capture, emotional arousal and behavioural prediction, children who already struggle with impulse control, transition, reward sensitivity, social judgement or emotional regulation may not simply be users. They may be unusually profitable users. That should disturb us.
There is now a growing body of evidence that adolescents with ADHD are more vulnerable to problematic social media use. A 2022 review argued that this vulnerability is likely shaped by impulsivity, reward sensitivity, emotion regulation, sleep disruption and social difficulties. Other studies have linked ADHD symptoms with problematic internet use and gaming, while recent work has explored problematic online gaming as a pathway between attention-deficit/hyperactivity symptoms and poorer mental health. (PMC)
This does not mean every child with ADHD will struggle online. It does not mean autism, ADHD or trauma automatically make a child unsafe. It does not mean parents are helpless or that children have no agency. But it does mean that the old advice — set limits, monitor use, teach resilience — is not enough.
A child who struggles to pause is growing up inside systems designed to remove pauses. That sentence should be at the centre of any serious conversation about neurodivergence and online safety.
The technology sector often talks about user choice. But choice is a weak word inside an environment engineered to reduce friction. The child does not choose each next video in the way we imagine a sovereign adult choosing from a shelf. The next video arrives. The game rewards. The notification interrupts. The shop flashes. The countdown begins. The feed refreshes. The algorithm learns. The child follows.
And over time, the system becomes more personal.
This is the part that makes the digital environment different from older forms of media. Television could be addictive, frightening, sexualised, violent or manipulative, but it did not adapt itself to the micro-patterns of a child’s attention in quite the same way. A magazine did not learn which image made a child pause. A playground rumour did not optimise itself around the child most likely to react. A slot machine did not follow a child into bed, into school, into the car, into every gap between one demand and the next.
A recommender system does not need to know a child has ADHD to learn from ADHD-like behaviour. It does not need a diagnostic label to notice quick clicking, late-night use, rapid switching, repeated viewing, intense engagement, angry commenting, compulsive returning, novelty seeking or attraction to high-arousal content. It only needs behaviour. Then it offers more of what holds interest.
That is not care. It is optimisation…
We have been through versions of this before. Slot machines were designed around intermittent reward. Fast food was engineered around salt, fat, sugar, speed and convenience. Supermarkets learned where to place sweets. Tobacco companies understood youth culture long before they were forced to admit what they were doing. Every era has its profitable vulnerabilities. What is different now is the intimacy of that extraction.
The phone sits beside the child’s pillow. It knows the hour they are most dysregulated. It knows what they watch after a fight, after exclusion, after shame, after boredom, after midnight. It knows which content they replay. It knows what makes them linger. It knows whether outrage holds them longer than calm. It knows whether fear, sex, violence, humiliation, status, conspiracy, self-harm, misogyny, gambling-style reward or fantasy identity keeps them there.
The platform may not “understand” any of this in a human sense. But it does not have to understand a child in order to shape their environment.
This is why the phrase “screen time” is so inadequate. It measures duration while missing design. One hour spent making music, messaging a trusted friend, watching a tutorial, joining a moderated special-interest community or playing cooperatively with known peers is not the same as one hour inside an algorithmic feed serving humiliation, violence, sexualised content, gambling-style rewards, influencer marketing and escalating provocation. The risk is then not just the time . It is pathway.
Language matters here. “Screen time” measures duration while missing design. “Addiction” can make the child look like the problem while leaving the environment intact. “Extremism”, “exploitation”, “self-harm”, “gaming”, “bullying”, “criminality” and “mental health” can each open one door while closing another. The wrong label can send a child into the wrong system. The narrow label can make adults miss the pathway. This is why emerging language such as “hybrid harms” and “weaponised loneliness” is useful, not because it gives us a perfect category, but because it helps us see what single labels often obscure: the overlap between design, vulnerability, belonging, coercion, status and harm.
Anger is part of this too. Not because “young people today” are uniquely angry, but because anger has become easier to detect, reward, amplify and convert into status. In some online spaces, humiliation, grievance and threat become social currency. The Resolver briefing describes Com-related ecosystems as clout-based environments where online infamy can function as a measure of status, with some groups even gamifying harmful acts through points, rankings and visible proof. Most young people will never go near those extremes. But the wider lesson matters: digital systems and subcultures can turn emotion into pathway, and pathway into identity.
Ofcom has recognised this in its online safety work. Its Children’s Register of Risks is intended to help services assess risks to children from harmful content, and Ofcom has repeatedly identified recommender systems as a major pathway through which children encounter harm online. In 2024, Ofcom said technology firms would need to “tame toxic algorithms” and ensure recommender systems do not operate in ways that harm children. (www.ofcom.org.uk)
This was a useful change. It moved the conversation from content moderation alone to system design. It recognised that harm is not only in the individual video, post, message or image. Harm can be in the route: what is recommended, what is amplified, what is repeated, what is made frictionless, what is pushed at the wrong moment to the wrong child. But even this may not go far enough for neurodivergent children.
Most online safety policy still treats “children” as if they are a single category. Sometimes age is considered. Sometimes content type is considered. Sometimes parental control is considered. Far less often do we ask how the same design feature might affect a child with ADHD differently from a neurotypical child, or an autistic child differently from a socially confident peer, or a traumatised child differently from a securely attached one. That is the missing layer.
A child with ADHD may be more vulnerable to novelty, reward, interruption and impulsive responding. An autistic child may be more vulnerable to literal interpretation, social manipulation, intense special-interest rabbit holes or difficulty detecting deception. A child with dyspraxia may already feel excluded from physical peer culture and find competence online. A traumatised child may be drawn to threat, control, fantasy, grievance or belonging. A child who is socially rejected at school may be more likely to trust the first online community that treats them as important.
These are not deficits in the child. They are points of fit between a child’s needs and a system’s incentives.
And when the incentives are commercial, fit becomes dangerous.
This is where the parenting discourse becomes morally lazy. Parents are told to monitor, restrict, supervise, educate, negotiate, install controls, check histories, use family agreements, remove devices at night and stay informed. Many do all of that. Some do it obsessively. Some are fighting devices, passwords, VPNs, multiple accounts, school tablets, gaming consoles, second-hand phones, peer pressure, app updates and the child’s own desperation to belong.
Parents are told to create boundaries inside an environment built to breach boundaries.
They are expected to be more persistent than the product designers, more technically skilled than the platforms, more available than the device, calmer than the algorithm, and more rewarding than the feed.
Then, when something goes wrong, the blame travels downwards: to the child who clicked, the parent who failed to monitor, the school that missed the sign, the safeguarding professional who did not join the dots.
The platform remains background. This is the pattern we have to break.
There are real risks here. Internet Matters advises that neurodivergent children are more likely to benefit from online spaces, but also more likely to experience harm; NSPCC and Ambitious about Autism have developed specific advice for families of children with SEND because generic online safety advice often misses the particular challenges these families face. (Internet Matters)
The challenge is that much of the advice still lands at family level. It tells parents and carers what to do after the design has already been deployed. That advice may be helpful, but it is not enough. It is like giving parents a leaflet on water safety while allowing companies to build faster currents around children who cannot yet swim.
The 5Rights Foundation has long argued that the digital world is risky by design. Its research using child avatars found that popular digital services could create automated pathways towards graphic self-harm images, extreme diet content, pornography and contact with adult strangers. Its broader work argues that commercial objectives become design features that shape children’s experiences. (5rights)
That phrase — risky by design — matters.
It means we should stop treating online harm as a series of unfortunate accidents. A child did not “just happen” to be drawn into an endless feed. A young person did not “just happen” to be served more extreme content. A game did not “just happen” to include reward loops, purchases, streaks, scarcity and status markers. A shopping app did not “just happen” to look like a game. These are design choices.
And design choices can be changed.
There is a political opening here. The UK’s Online Safety Act has begun to shift duties onto platforms, and Ofcom’s child safety regime now requires services likely to be accessed by children to assess risks and put protections in place. In 2026, Ofcom also issued a call for evidence for its first statutory report on content harmful to children, due by October 2026. At the same time, campaigners continue to argue that the regulatory regime must go further on addictive and manipulative design features, not only illegal or obviously harmful content. (www.ofcom.org.uk)
This matters because the next phase of online safety cannot only be about taking down the worst content after children have already encountered it. It has to ask why children were routed there, why they stayed there, and whether the design itself increased the likelihood of harm.
For neurodivergent children, that question is urgent.
The child who struggles with transitions may find it unbearable to stop when the next video has already started. The child who is intensely reward-sensitive may be pulled into streaks, points, levels, drops, rankings, skins, loot boxes or surprise rewards. The child who is socially anxious may find online spaces safer at first, then become dependent on them. The child who is bullied may find belonging in communities that harden grievance. The child who is impulsive may send the message, image or threat before the future has arrived in their mind. The child who is literal may not spot manipulation. The child who is lonely may mistake attention for care.
And the traumatised child may return, again and again, to material that feels familiar not because it is safe, but because their nervous system already knows fear.
This is where the recent Resolver briefing, Weaponised Loneliness, is useful. Resolver uses the phrase not to suggest that loneliness alone causes harm, but to describe the deliberate exploitation of isolation and related vulnerabilities as tactical threat vectors within a wider online harm ecosystem. In its account of the Com, loneliness is both a pathway into harmful participation and a point of leverage used by predatory members. The report is not about ordinary screen use, and it would be wrong to collapse all online risk into its most extreme forms. But it does name something this debate often misses: isolation, status, belonging and vulnerability can be engineered around, recruited through and exploited.
That matters for neurodivergent children because the risk is not only exposure to “bad content”. It is the fit between a child’s unmet need and a system that can identify, amplify and monetise attention. A child who is lonely may not experience a message, invite, gaming reward, group chat or algorithmic recommendation as a risk signal. They may experience it as recognition.
None of this is simple. Digital spaces can also be where neurodivergent children find identity, creativity, friendship and expertise. Some young people learn more from YouTube than from school. Some find their first real community through gaming. Some autistic children communicate more freely online. Some ADHD children make, remix, perform, joke, research, compose, edit, build and connect with extraordinary intensity. A safer internet cannot mean removing neurodivergent children from digital life.
It has to mean changing the terms on which digital life meets them.
That means moving beyond “parental controls” towards design duties.
A neurodivergent-safe digital environment would not rely on a child’s self-control as the main safety mechanism. It would not assume that every user can ignore notifications, resist autoplay, interpret social risk, manage spending prompts, detect manipulation, pause before sharing, or leave when distressed. It would build friction where friction protects agency.
It would allow stronger default settings for children and vulnerable users. It would make safety features harder to bypass. It would limit autoplay, infinite scroll, late-night nudges and manipulative notifications for under-18s. It would reduce algorithmic escalation towards high-arousal or harmful content. It would provide meaningful recommender controls, not decorative settings buried in menus. It would treat gambling-style mechanics and spending prompts as safety risks, not just consumer choices. It would recognise that some children need systems that help them stop.
This is not about making digital life dull. It is about refusing to build childhood around compulsion. It is also about justice.
The children most likely to be blamed for their online behaviour may be the least protected by the current model. Neurodivergent children are often already navigating school systems that misread distress as defiance, health systems with long waits, families under pressure, and peer worlds where difference can be punished. If the digital environment then amplifies impulsivity, isolation, sensation-seeking or shame, the harm does not remain online. It travels back into the home, the classroom, the playground, the safeguarding referral, the police record, the child’s sense of self.
This is why Safe by Default matters.
It is not enough to teach children resilience inside systems designed to overpower it. It is not enough to ask parents to supervise every click. It is not enough to tell schools to confiscate phones while the wider digital environment remains built around extraction. It is not enough to treat neurodivergent children as edge cases, when they may be among the clearest indicators of whether a system is humane.
The test of a digital environment is not how it works for the calmest, safest, most resourced child on their best day.
The test is how it works for the child who is tired, impulsive, lonely, ashamed, dysregulated, excluded, grieving, autistic, ADHD, traumatised, sensation-seeking, socially naïve, desperate to belong, or already known to adults as “too much”.
If the system is safe for that child, it is likely to be safer for everyone.
If it is dangerous for that child, we should not call the resulting harm a parenting failure. We should call it a design failure. And then we should redesign it.
Source notes
Dekkers et al., 2022 — problematic social media use and ADHD
Adolescents with ADHD are more vulnerable to problematic social media use, with likely mechanisms including impulsivity, reward sensitivity, social difficulties, emotion regulation and sleep disruption. (PMC)
Ghiaccio et al., 2025 / broader ADHD and problematic internet use research
ADHD and problematic internet use, including gaming-related harms. (PMC)
Narita et al., 2025, Nature Mental Health — problematic online gaming
Attention-deficit/hyperactivity symptoms and problematic online gaming can interact with mental health outcomes. (Nature)
Ofcom Children’s Register of Risks / Online Safety Act child safety duties
Services must assess and mitigate risks to children from harmful online content. (www.ofcom.org.uk)
Ofcom on recommender systems
Recommender systems can be a major pathway to harm and that platforms need to ensure algorithms do not harm children. (www.ofcom.org.uk)
5Rights Foundation — risky by design / Pathways avatar research
Commercial objectives become design features that can route children towards harmful material. (5rights)
Internet Matters / NSPCC and Ambitious about Autism SEND online safety advice
Neurodivergent children can benefit from online spaces but may also face specific risks requiring tailored advice and support. (Internet Matters)
Recent news : NSPCC sextortion / online sexual exploitation rise
The Guardian reported a sharp rise in Childline counselling sessions about online sexual abuse and exploitation, particularly blackmail involving children’s sexual images. (theguardian.com)
Resolver, 2026, Weaponised Loneliness: A Critical Harm Intelligence Briefing
Hybrid online harms, the Com, nihilistic violent extremism, sadistic exploitation, clout-based online subcultures and the exploitation of isolation, vulnerability and social need. Especially relevant to the argument that online harm is not only about content, but about pathways, belonging, status, grooming, platform features and the speed at which harm can escalate.