Beyond the plea bargain: Open questions on Meta and children’s mental health

The share price rose on the stock market. The market took it well. Statements from US attorneys-general presented the settlement with Meta as a historic turning point: billions of dollars and new rules to protect children. The settlement brings the legal dispute to a close without resolving the very question that had made it possible in the first place: whether, to what extent, and through what mechanisms these platforms cause harm to children’s mental health. It is worth remembering that this was the crux of the matter: not a vaguely defined sense of unease, but actual harm to mental health.

An agreement may be politically satisfactory, economically rational and prudent. And yet it need not amount to an admission of guilt. Meta has not admitted liability, and the case has not reached a final determination in which evidence, expert testimony and cross-examination were thoroughly assessed. What prevailed was a strategy of damage limitation on the part of the company and a cautious approach on the part of the opposing party. A remedy was found before the nature of the problem had been established with sufficient precision.

The structure of the measures illustrates this well. A daily limit of two hours, a night-time block, restrictions on ‘likes’, filters and engagement mechanisms may be based on a cautious approach, but they do not automatically stem from a scientifically established dose-response relationship. Why should two hours be a safety threshold? What harm occurs from the third hour onwards? No one has demonstrated this. Social media is not like the old local TV channels after the ‘child protection’ hours ended: the idea that risks only begin at midnight and end at six is a curious hypothesis.

Obviously, public policies are full of conventional constraints introduced in the face of uncertainty. Conventions do not need to be ‘right’ but simply to work – that is, to be accepted as conventions. A certain threshold may be adopted simply because it reassures public opinion. That is no small matter. The risk, however, is confusing this sort of practice with the solution to a problem whose existence should have been established by the proceedings.

The case of ‘likes’ is emblematic. The public visibility of ‘likes’ associated with specific content should be restricted. This is a solution which, in Instagram’s internal trials, had shown limited effects on negative social comparisons. It is now included in the package of measures for minors. Will it bring benefits for those users? Or will the effects prove too modest? During the process leading up to the agreement, there was lengthy debate over whether the involvement of the dopaminergic system was sufficient to demonstrate a specific mechanism of addiction linked to ‘likes’. Dopamine is released during any rewarding experience, from food to music to gaming, and to say that a piece of content ‘triggers dopamine’ is no more informative than saying it triggers a heart rate. Removing the ‘likes’ counter addresses a plausible symptom, not a proven mechanism.

The settlement provides for mechanisms to verify outcomes, but in any research, it is the questions that matter more than the answers. What is missing – at least in the public framework of the agreement – are pre-defined endpoints, control groups designed to isolate the effects, and a framework that allows us to distinguish the improvement brought about by the measures from trends that were already underway. It is possible that spending less time on mobile phones could have positive effects for young people. If we side with conventional wisdom, this is undoubtedly the case. It is likely that if less time on social media equates to more time spent outdoors with peers, there may, for example, be improvements in terms of mood and interpersonal skills. Would this confirm the accusatory narrative?

A policy is not a medicine. But if its justification is framed in terms of mental health, the absence of a robust evaluation strategy becomes more, not less, significant.

To introduce a restriction, it is sufficient for a risk to be socially credible and politically urgent. To remove it, however, one must either demonstrate that it is unnecessary, or bear the political cost of appearing less protective towards children. This is why we are surrounded by anachronistic regulations that have outlived the problems they were intended to solve.

The agreement is perfectly logical. The attorneys general secure funding, visible measures and a result that can be communicated politically. Meta reduces a huge litigation risk, stabilises part of its financial exposure and helps to set a standard which, if adopted by competitors, can spread compliance costs across the entire sector. Both parties have rational reasons for preferring a settlement to a court ruling.

This is the paradox of mass litigation. When the financial risk becomes high enough, a settlement can become a rational option even for a party convinced it can win on the merits. Payment does not measure fault: it measures the price of uncertainty. For the State, in turn, a very substantial settlement may be preferable to a future but uncertain victory, especially when it allows it to immediately secure measures that meet a strong political demand. It may therefore be in both parties’ interests to settle the very case that could have clarified the underlying issue.

The effect is amplified when the financial aspect of the agreement is transformed, in public discourse, into an epistemic shortcut: they paid a lot, so the claim must have been true.

One wonders what the implications will be for the regulation of complex technologies. The EU has already asked Meta to adapt its practices in Europe as well. It is understandable that precautionary rules should be adopted, particularly when children are involved. However, precaution must be distinguished from the claim that the mechanism of harm has already been proven, and should incorporate mechanisms capable of adjusting in the light of the results.

Care should be taken to ensure that the political weight of an agreement is not mistaken for the epistemic weight of evidence. The deal met the most immediate expectations of politics and the market: reducing uncertainty, demonstrating action, spreading the costs, and resolving part of the conflict. But the most difficult question remains unanswered. Until it is treated as an empirical question, we will continue to know with great precision what we want to ban without knowing with the same precision why.