Facebook, depression and billions of dollars: what we actually know

How much evidence do we need before asking a social network to protect the children who use it?

We are still arguing over the causes of depression and anxiety across a whole generation. Meanwhile, with Meta we are already talking about billions of dollars, internal warnings and product changes. That difference is precisely what interests me. In debates about technology, we have grown used to waiting for one definitive answer to decide whether social networks are good or bad. But a child receiving unsolicited sexual messages cannot wait for the outcome of that debate.

And in my view, a company that knows about such experiences cannot hide behind it.

Seventeen billion. And what does it buy?

On 26 August 2026, California Attorney General Rob Bonta announced a proposed settlement with Meta. According to his office, the company is to pay states up to 17 billion dollars over ten years. Headlines also refer to approximately 18 billion; I am sticking to the amount and scope of the official announcement.

It is tempting to call it a fine. But it is a proposed agreement subject to court approval, and Meta denies wrongdoing within it. The amount itself therefore does not prove that Instagram caused anyone's depression. Companies also settle because of the costs and uncertainty of litigation.

More interesting to me than the number of zeros are the changes Meta accepted. The proposal includes a default two-hour daily limit for users under eighteen and blocking use from midnight to six in the morning, with the possibility of a parental override. It also includes notification restrictions, a ban on displaying like counts to teenagers, the option of a non-personalised feed and an independent audit. We are discussing an American agreement, not rules that automatically already apply to Czech children.

Suddenly, then, the product can be changed too. The issue is not only whether parents should monitor phones more consistently. It is how the app is built, what defaults it offers users and what it encourages them to do.

A parent should raise their child. They should not have to be the only counterweight to a system designed to keep that child inside.

Science has not said ‘nothing is happening’

‘We do not have scientific confirmation that social media causes depression and anxiety.’ I understand that sentence, but today I would not write it so broadly. We already have some evidence of causal effects. The main dispute concerns their extent and what we can infer for different people and platforms.

In a report published around the turn of 2023 and 2024, the US National Academies concluded that the literature reviewed did not support a conclusion of causal changes to adolescent health at population level. Put simply: it is insufficient to explain a whole generation's health changes through social media. The report nevertheless describes specific risks and benefits.

Alongside this, we have the work of Braghieri, Levy and Makarin, published in 2022 in the American Economic Review. They made use of Facebook's gradual rollout at American universities. Their quasi-experimental analysis found a negative effect on students' mental health. The method seeks to separate causation from mere coincidence, but relies on certain assumptions. And it examines early Facebook among university students. A thirteen-year-old Instagram user today is a different situation.

Then there are experiments in which use actually changed. In a 2025 NBER working paper, Allcott and colleagues compared users who deactivated Facebook or Instagram for six weeks before the 2020 US elections with a control group that switched them off for the first week only.

The longer disconnection produced a small average improvement in an index of happiness, depressive feelings and anxiety: 0.060 standard deviations for Facebook and 0.041 for Instagram. These are effect sizes, not percentages of people cured. The authors measured emotional state during a particular period, not the onset of clinical depression in children. The version cited is a working paper, and the co-authors include Meta employees.

So yes, in certain conditions we already have evidence of causal negative effects. But it will not give us a single number explaining all depression, all anxiety and all children. If only it were that simple — we would spare ourselves a lot of bad headlines.

Two hours of what?

One teenager spends the evening talking to a friend. Another posts a drawing and receives their first encouraging responses. A third scrolls through content that makes them increasingly disgusted by their own body. They may all end up on the same line in a screen-time table.

That is precisely why ‘how many hours?’ is not enough for me on its own. What happens during those hours? Whom is the person in contact with? What does use displace from their day, and what were they already bringing into it?

Loneliness may lead someone to a network. There, they may find people they belong with, but also further rejection. Anxiety may lead to repeated phone-checking, while night-time use can worsen sleep. Add family, school, bullying and previous difficulties. Two rising curves will not untangle that knot.

And the average? It can be small even when a platform helps some and seriously harms others. The average user will not tell you what is happening to your particular child.

What people inside knew

In 2021, Frances Haugen released internal Facebook materials and testified before the US Senate. The company studied young Instagram users' negative experiences, including difficulties with body image. Facebook denied withholding information at the time and emphasised that its platforms helped many people.

Certainly, a user's answer that ‘Instagram makes my difficulties worse’ does not establish causation by itself. But it is information that product management should work with. Imagine users repeatedly describing a harmful experience to you. Do you really need to solve the whole population's mental health first before beginning to address it?

Former safety engineer Arturo Béjar also testified about this at the August 2026 trial. According to reporting on his testimony, he warned leadership about unsolicited sexual contacts, young users' harmful experiences and shortcomings in reporting. He said he sent warnings to Mark Zuckerberg too. In his view, the company's public statements created a misleading impression of its protection of children. Meta rejected the allegations.

Here I ask about responsibility. Béjar's testimony is testimony, not automatically a court-confirmed account of every event. But it opens very specific questions: what warnings reached leadership, who assessed them and what changed afterwards?

Uncertainty about the causes of a whole generation's depression does not remove the obligation to address the harassment of a particular child.

Who should take the next step

Follow the money. With a platform that earns from advertising, it is worth asking how much weight user time carries in decisions and how much their experience carries. That alone proves no particular wrongdoing. But it explains why we cannot be satisfied with assurances that the company takes safety seriously.

I want to see reporting systems that work. Safer defaults. Independent assessment of whether measures help. Researcher access to data that can reveal more than screen time.

At home and at school, I would start much more ordinarily. ‘What makes you feel worse on your phone?’ ‘Who writes to you there?’ ‘What would you miss if you didn't have that app?’ Perhaps we will learn about a relationship we do not want to take away from the child. Perhaps about something they need help with. Without that, we can easily ban a source of support and overlook the real problem.

And yes, time matters, especially when it eats into sleep. But let us not make it a substitute for interest in what the child is living through.

I would want the same from Meta at the scale for which it is responsible. Find out what is happening to users and act accordingly. We may wait a long time for a complete explanation of depression. We must not wait that long for effective help for a child who reports harassment.