← Symptoms

Mental Health

The evidence is extensive, contested in detail, and clear in direction. Here is what is established, what is genuinely uncertain, and what Facebook knew.

How to read this page. Tap any underlined word to see the precise academic term and a short definition. Expand any "Deeper" box for the evidence and contested points. The main text works on its own — you can skip both and still get the whole argument.

What is well-established

Adolescents who spend more than three hours per day on social media are significantly more likely to report anxiety and depression symptoms — a US cohort study of 6,595 teenagers found the odds of internalising problems (anxiety and depression symptoms) about 60% higher at three to six hours a day than with no use, and roughly double for internalising and externalising problems together. A large UK cohort study finds the association stronger for girls than for boys. These findings appear across multiple large studies in the US, UK, and internationally; the correlation itself is not seriously disputed. Some studies find passive scrolling — consumption without active social interaction — linked more strongly to negative outcomes than active social use, but that finding is contested: a 2022 review of 40 survey studies found that most did not bear it out.

The timing is the most argued part of the case. On the measures Jonathan Haidt assembles, adolescent mental health in the US, UK, and other high-income countries began deteriorating around 2012 — the years in which smartphones spread to most American teenagers (37% had one in 2012 and 73% had access to one by early 2015, in Pew Research Center's surveys) and social media became the dominant leisure activity for teenagers. His synthesis in The Anxious Generation (2024) documents the convergence across data sources and countries; his critics, below, argue that the convergence is weaker than it looks, and that the timing does not establish the cause.

What Facebook knew

In 2021, Frances Haugen disclosed internal Facebook research documents to the Wall Street Journal and to Congress. The documents showed that Facebook's own researchers had found that, among teenage girls who already felt bad about their bodies, one in three said Instagram made them feel worse — the slide's own headline was "We make body image issues worse for one in three teen girls." A 2019 internal study found that among teenagers who reported suicidal ideation, 13% in the UK traced the desire to kill themselves to Instagram. The company knew. The product continued. Meta's reply was that the research was meant for internal discussion, that some of it rested on as few as 40 teenagers, and that it did not measure causal relationships. All of that is true, and none of it touches the point: the slides matter as evidence of what the company knew, not as a clinical study.

The Haugen documents are primary source evidence — not the allegation of a critic but the company's own research conclusions. They show the findings being discussed internally and played down in public, and the fixes the researchers proposed not being adopted at scale.

The debate: Haidt vs Orben

The scientific debate on causality is genuine. Amy Orben and Andrew Przybylski at the Oxford Internet Institute have published influential critiques arguing that the effect sizes in most studies are small and that the correlational evidence does not establish causation. Their 2019 paper found that the association between adolescents' digital technology use (social media included) and wellbeing was comparable in size to that of wearing glasses or eating potatoes.

Jonathan Haidt and Jean Twenge argue these critiques understate the evidence and that the methodological choices Orben and Przybylski made minimised effect sizes. The dispute is ongoing in the academic literature: Candice Odgers' review of The Anxious Generation in Nature argues that the timing evidence is weaker than the book presents, and that other causes of the decline remain in play. On the critics' side, a 2022 meta-analysis by Christopher Ferguson and colleagues of 33 studies found only small associations between screen media and mental health. On the other, the two studies closest to an experiment both point the same way: a randomised four-week Facebook deactivation improved a wellbeing index by 0.09 standard deviations, and Facebook's staggered arrival at US colleges was followed by a 0.085 standard-deviation rise in poor mental health — small effects, on adults, but in the direction the correlations suggest.

The honest position, which this site holds: correlation between heavy social media use and worse mental health outcomes is established across multiple studies. Causality is contested. The direction of the evidence is clear enough to act on, particularly given what the internal platform documents show about knowledge of harm.

The sleep pathway

One route from heavy use to worse mental health is better documented than the rest: sleep. A meta-analysis of 20 studies covering 125,198 children found that using a phone or tablet at bedtime was associated with too little sleep (odds ratio 2.17), poorer sleep and daytime sleepiness — and that a device merely present in the bedroom, even unused, was associated with the same (odds ratio 1.79 for too little sleep). A systematic review of 42 studies of 16-to-25-year-olds found the same three-way pattern: heavy social media use with poor sleep, and poor sleep with anxiety and depression, which is why sleep is the most plausible mediator. These are associations in observational studies, not proof of cause; but they are consistent, large, and they point at a single household-level change — the phone out of the bedroom at night — that costs nothing to try.

Why "not addiction, design"

Framing heavy social media use as "addiction" is usually inaccurate, and it is counterproductive. A 2025 study in Scientific Reports found that Instagram users who saw their heavy use as an addiction — rather than as a habit — felt less able to control it, that most who called themselves addicted did not meet clinical criteria, and that loose use of the word by media and institutions feeds the overestimate. The accurate framing is: the product was designed to be this compelling. That reframe preserves the user's capacity to respond.

How we know — why the correlation is solid but causation stays contested +

Most of the evidence is cross-sectional: it measures social media use and wellbeing at one moment and finds them associated. That cannot tell you the direction — whether the feed worsens mood, or low mood drives more scrolling, or a third factor drives both. This is the core of Orben and Przybylski's critique. Their widely-cited analysis used a "specification curve," running thousands of defensible model variants to show the average measured association is small and sensitive to analytic choices.

What would settle it, and why it's hard. Stronger designs — large longitudinal cohorts that follow the same children for years, and randomised reductions in use — exist for adults (the deactivation experiment and the college roll-out study above) and are slow to arrive for children, because randomising children's social media for long periods is ethically and practically fraught. So the debate persists honestly.

Why we still say "act on it." Two things tip the balance without resolving the academic question: the timing evidence across countries (Haidt's synthesis), and the Haugen documents. Facebook's own surveys do not settle the science; they are no stronger than the published studies. What they show is conduct: the company saw the signal and the product continued. No effect-size debate touches that, and it is what moves this from "interesting correlation" to "grounds for action."

What you can do →
Where to get help →
Body Image →
For Parents →
Yes, but… →