Jonathan Haidt, in The Anxious Generation, argues that the arrival of smartphones and social media between roughly 2010 and 2015 transformed childhood from play-based to phone-based and thereby caused a sharp rise in anxiety, depression and self-harm among teenagers, especially girls. Candice Odgers, reviewing the book in Nature, accepts that young people are “in crisis” and that platforms need reform, but insists Haidt’s causal claim is not supported by the evidence. Both positions have merit: Haidt offers a coherent causal hypothesis, and Odgers is right that much of the evidence is correlational and therefore insufficient on its own to establish causation. The task is to weigh plausibility against proof and to propose what policy should follow from that balance.
The statistical evidence is the first battleground. Large-scale surveys show a rise in adolescent mental-health problems beginning around 2012, coinciding with the rapid spread of smartphones and social apps. Haidt points to similar timing across Canada, Britain, Australia, New Zealand and the Nordic countries, arguing that a cross-national pattern makes purely American explanations unlikely. Critics counter that associations between “screen time” and well-being are tiny. Analyses of hundreds of thousands of adolescents have found that digital-technology use explains at most a fraction of a per cent of the variance in well-being, while other measured factors, such as being bullied, show much larger associations. That suggests effect sizes for technology, taken as a broad category, are small.
But the devil is in the measurement. “Technology use” often lumps together television, gaming and social media, and averages across boys and girls; disaggregate the data and the picture can change. Haidt and colleagues argue that the links with distress are stronger for image-based social platforms and for girls. The same dataset, parsed differently, can therefore support different conclusions. This is not a quibble; it is a methodological point: cross-sectional correlations are sensitive to how variables are defined and to unmeasured confounders.
History counsels caution. New media, from novels and radio to comics, television and video games, have repeatedly been blamed for youth problems, only for those panics to recede as evidence accumulated. That pattern should temper alarm. Yet past false alarms do not prove that smartphones are harmless. Two features of smartphones and social media seem qualitatively different from earlier media: phones are portable and intimate, following teenagers into bedrooms and sleep cycles, and social media quantifies popularity, turning peer status into a public metric. Those differences make Haidt’s displacement and comparison mechanisms plausible, but plausibility is not proof.
The central empirical challenge is distinguishing cause from coincidence. Two broad explanations fit the rising rates of distress. One: smartphones and social media caused the decline. Two: smartphones arrived alongside other social changes, such as economic hardship after 2008, shifting family dynamics and educational pressures, and the decline stems from those factors, or from reverse causation, in which struggling teenagers gravitate to online spaces. Longitudinal studies sometimes find evidence of reverse causation: adolescents with pre-existing difficulties go on to increase their social-media use. Time-trend graphs in which both lines rise together cannot adjudicate between these accounts, because both predict such patterns.
What would settle the question is causal evidence of the right kind. Natural experiments that exploit staggered smartphone adoption across places, or randomised trials that remove access to social media, would be far more informative. Haidt points to experiments in which participants who quit social media report improved well-being; Odgers cites cross-country analyses that find no consistent link between the spread of social platforms and mental-health trends. Both lines of evidence are suggestive but incomplete: short-term experimental effects may not generalise to population-level trends, and ecological cross-country studies can mask subgroup effects.
The policy stakes are real and asymmetric. If Haidt is right and we delay intervening while demanding perfect proof, a generation may suffer avoidable harm. If Odgers is right and we act prematurely, resources and attention may be diverted from more important causes, and some adolescents, particularly those who are isolated or marginalised, may lose valuable online support. The right public response should therefore be risk-sensitive: prioritise measures that are cheap, reversible and designed to generate causal evidence.
Haidt’s policy proposals (no smartphones before high school, no social media before sixteen, phone-free schools) are community-level interventions intended to avoid the social exclusion that would follow unilateral parental bans. They are also testable. A pragmatic approach is to pilot phone-free school days or phone-restriction policies in a trial in which schools are, where feasible, randomly assigned to different phone policies, with pre-registered outcome measures (sleep, anxiety, depression, social connectedness). Such policies are easily reversed, and a well-designed trial could produce the longitudinal and comparative data the debate lacks.
In sum, Haidt supplies a powerful, coherent causal hypothesis grounded in plausible mechanisms; Odgers supplies a necessary methodological corrective, reminding us that correlation is not causation. The evidence to date is mixed: small average associations in broad surveys, stronger signals in some disaggregated analyses, and conflicting results from experiments and cross-national studies. The most honest position is therefore provisional: the deterioration in adolescent mental health is real; its causes remain uncertain; and policy should favour proportionate, reversible interventions that can also generate better causal evidence. Trial, measure and revise, until we can say with confidence whether phones are a primary cause, a contributing factor, or largely a symptom of wider social changes.