We keep hearing that digital technology is making young people more stupid. In 2026, Finnish researchers published the results of eight years of measurements that run counter to this judgement, and recent research, read closely, is far more nuanced than the ambient panic would suggest. I reread these findings with Michel Serres, and draw a proposal from them: to grant young people a presumption of competence, including in the face of artificial intelligence.
“Young people can’t concentrate any more”
In the training sessions I lead, in meetings of educational and cultural teams, one sentence keeps coming back in various forms: young people can no longer concentrate, social media is making them stupid, artificial intelligence will finish the job. It is uttered as a shared truism, calling for neither evidence nor discussion. In Youth as a power relationship I developed the idea that judgements of this kind tell us less about young people than about the position of the person making them, since declaring a generation incapable spares you having to listen to it and leaves the adult with the last word. And I recalled in Rethinking youth culture that the complaint about the decadence of the young is as old as writing itself: we find it already in Hesiod, in the eighth century BCE.
Here I want to examine a narrower question. When people claim that screens and social media are degrading young people’s cognitive abilities, they readily invoke science. Yet there now exist large studies that have followed the same children over many years, with serious measuring instruments, and their results do not point where they are said to point.
Eight years of measurements in Finland
In 2026, Petri Jalanko, a doctoral researcher at the University of Jyväskylä, published with five colleagues, in the journal Pediatric Exercise Science, the results of an eight-year follow-up conducted within the PANIC cohort (Physical Activity and Nutrition in Children), run by the universities of Jyväskylä and Eastern Finland (Petri Jalanko, Marja Leppänen, Bert Bond, Jari Laukkanen, Timo Lakka and Eero Haapala, “Associations of Physical Activity and Sedentary Time From Childhood to Adolescence With Cognition in Adolescence: The PANIC Study”, Pediatric Exercise Science, vol. 38, no. 3, 2026). 260 adolescents, 124 girls and 136 boys, with a mean age of 15.8 at the time of the final tests, had been followed since childhood. Screen time and physical activity were measured both by questionnaires and by a sensor combining heart rate and movement, worn for several days at a stretch. Working memory, attention and learning were assessed with a battery of standardised tests.
The researchers write that their first result took them by surprise: the more screen time accumulated since childhood, the better the overall cognitive score in adolescence. The picture is not uniform, since the same screen time went along with slower reaction times on some working-memory tests, and time spent sitting away from screens was, for its part, associated with poorer performance. The team also notes that the link between physical activity and cognition depends on sex and on the type of activity. Nothing in these data allows the conclusion that “screens make you smarter”. But nothing, and this is the point that matters to me, supports the opposite thesis either, the one presented to us as settled.
The study’s authors are extremely cautious, and their caution is methodologically sound. Cognition was measured only once, in adolescence, which rules out establishing cause and effect, and the causality could very well run the other way: it would not be screens improving cognition, but adolescents with the most solid cognitive development organising their time, screens included, differently. The team itself raises this hypothesis about physical activity, and the reasoning holds for all of its results. That is why I refrain from making this study say that screens make you smarter, which would amount to manufacturing a reverse causality. I find something else in it, and that is already a great deal: those who are prosecuting the dumbing-down case do not have the evidence they lay claim to, and to understand cognitive development you have to start from people, their living conditions and their environment, rather than from the screen as an object. Petri Jalanko himself goes no further than advising parents and teachers to steer usage towards active thinking, problem-solving, creation and learning rather than passive consumption.
The so-called marshmallow experiment is a now-classic example of manufactured causality. In the 1960s and 1970s, Walter Mischel and his team at Stanford offered young children one treat straight away, or two if they waited; the children who had managed to wait showed, years later, better school careers, and a whole pedagogy of waiting was built on this. In 2018, Tyler Watts, Greg Duncan and Haonan Quan repeated the protocol with more than 900 children from a far more representative cohort: the correlation was half as strong as in the original studies, and it shrank by a further two thirds once family background, early cognitive abilities and the home environment were taken into account (“Revisiting the Marshmallow Test”, Psychological Science, vol. 29, 2018). Most of what the little test seemed to predict came down to the children’s social background. For half a century, a correlation had been taken for a cause, because it confirmed what people wanted to prove.
With this caution in place, the asymmetry in how these results are treated strikes me. When researchers find, on far more fragile data, a negative association between screens and cognition, the press turns it into sweeping headlines and policy-makers turn it into public policy. When eight years of sensor-based follow-up end in a positive association, the conclusion never gets past the laboratory door, hedged about with caveats. I am not asking for sweeping headlines in the other direction, which would manufacture an equally unfounded causality; I am asking that the same restraint apply both ways.
In France, the Haut Conseil de la santé publique takes a position close to that of the Finnish researchers. In its opinion of 12 December 2019 on the effects of children’s and young people’s exposure to screens, supplemented on 8 March 2021 by a second opinion on excessive use, it notes that the scientific literature offers contradictory findings on children’s cognitive development, that these uncertainties call for more research, and that the real issue lies in training, education and the guidance of usage. In Cloud of Nuanced Screens I proposed a grid for stepping out of the panic, by distinguishing the medium, the content and the use.
Yann Leroux sifts the evidence
The psychologist and psychoanalyst Yann Leroux published in August 2026 a literature review, “L’attention est-elle vraiment détruite ?” (“Is attention really being destroyed?”), in which he goes back over ten experimental studies published between 2023 and 2026 to disentangle what is solidly established from what still belongs to intuition. What emerges is not a grand narrative of cognitive collapse but a series of distinct mechanisms, each with its own level of evidence. Among the results he counts as solid:
- a notification causes a transient cognitive slowdown of about seven seconds, and this cost comes from the repeated fragmentation of activity rather than from accumulated hours of screen time (Fournier et al., Computers in Human Behavior, 2026);
- algorithmic personalisation keeps usage going and makes it harder to stop: in an experiment comparing a regular TikTok feed with a depersonalised one, switching to the depersonalised feed cut usage time by about forty minutes a day (Dekker, Baumgartner and Sumter, Telematics and Informatics, 2025);
- infinite scrolling has a cost for memory, and people remember content better when the interface slows the scrolling down a little (Ruiz, Molina León and Heuer, 2024).
These effects are real and measured, sometimes with random assignment of participants. But they stem from the design of interfaces, which was devised by people and which other people can change, debate and regulate. The thesis of a lasting, across-the-board degradation of young people’s attention is of an entirely different nature, and it remains fragile. Studies that follow the same people over time are scarce, the effects small, and most of the links observed are mere correlations. A meta-analysis published in Psychological Bulletin, covering 71 studies and 98,299 participants, finds links between heavy use of short videos and poorer attention, but the studies diverge widely from one another and most capture a single moment in time, which establishes a correlation, not a cause (Lan Nguyen et al., “Feeds, Feelings, and Focus”, Psychological Bulletin, vol. 151, no. 9, 2025). Yann Leroux concludes that the right question is not whether screens are destroying attention, but which mechanism is at work, at what moment, for how long, and with what concrete consequences in everyday life. He sums it up in a formula: “the interfaces are new, the brains are unchanged”.
This sifting echoes older work. In 2019, Amy Orben and Andrew Przybylski analysed in Nature Human Behaviour data from 355,358 British and American adolescents, systematically testing tens of thousands of ways of analysing the same data in order to neutralise the biases of each research team. Digital technology use explained at most 0.4% of the variation in adolescents’ well-being, an effect of the same order as regularly eating potatoes. And in 2024, the psychology researcher Candice Odgers wrote in Nature, about Jonathan Haidt’s best-seller The Anxious Generation, that the thesis of a “rewiring” of children’s brains by digital technology is not backed by the science, almost all the data invoked being mere correlations. This debate among researchers is lively, and it is not settled. Its very existence shows that the dumbing-down case, presented to the general public as an established fact, has the status of a contested hypothesis within research itself. In Attention Reinvented I analysed what the discourse of the “attention crisis” conceals, namely the emergence of a new demand among audiences who have become creators.
Petite Poucette, head in hand
Michel Serres had thought all this through as a philosopher, fifteen years before these measurements. In Petite Poucette (Le Pommier, 2012), he places digital technology in the continuation of two earlier revolutions of the medium-message pair, the shift from the oral to the written, then from the written to print. Each of these revolutions externalised part of our human faculties onto a material medium, and each prompted the same complaint. Socrates, in Plato’s Phaedrus, already reproached writing with destroying memory, since no one would retain what they can look up. The same worry was heard when printing arrived. Serres turns the argument around: it is because the printed book relieved minds of the burden of memorisation that Montaigne could prefer a well-made head to a well-filled one. Externalisation does not empty the head, it frees it for something else, which Serres calls invention.
The image Serres gives of Petite Poucette, this young woman writing at full speed with her thumbs, is that of Saint Denis. She holds her cognition in front of her, in her hand, “as Saint Denis holds his head away from his neck”: memory, imagination, part of reason are now objectified in the computer and the phone. This emptiness frightens those who see it as a loss; Serres sees in it the condition of an intelligence made available for what no machine does in our place, asking questions, inventing, connecting. I extended this reading in Petite Poucette 2.0, where I propose to see in social media the new printing press, that is, a revolution in access to expression that the powers that be have every reason to demonise.
Recognising this generation’s new skills, its ability to navigate, to sort, to moderate its own use, in no way prevents us from working on the capture mechanisms described above. It is what makes that work possible, because we then stop treating young people as passive victims and start considering them as interlocutors capable of understanding how an interface was designed and of acting on their own practices.
Artificial intelligence on trial in its turn
The same pattern is playing out today around artificial intelligence. We are told it will make us stupid, that it will cause our abilities to write, to reason, to search to wither away. The centrepiece of the case so far is a preprint from the MIT Media Lab, “Your Brain on ChatGPT” (Nataliya Kosmyna et al., arXiv, June 2025): 54 people wrote essays under electroencephalography, and those writing with an artificial-intelligence assistant showed lower brain connectivity during the task. This study, a short-term measurement on a single task, has not yet been peer-reviewed, and the team itself asks that its conclusions be treated as preliminary; it has publicly voiced concern at headlines that had it saying artificial intelligence makes people stupid, wording it uses nowhere. The extrapolation is the same as with screens: from a local measurement, a diagnosis is drawn of a general degradation of our abilities, which nothing establishes at that scale. Yet we have an instructive precedent. With the printing press, the text, and interpretation with it, passed into the hands of people who had been excluded from it, and it was this shift in power that alarmed the institutions holding the monopoly on legitimate knowledge, to the point where they censored, controlled translations and drew up indexes of forbidden books. The authorities of the time did not fight the book because it made people stupid, but because it emancipated them.
My hypothesis is that artificial intelligence belongs to the same logic. It does not replace our intelligence, it shifts where that intelligence is exercised and extends its reach: more people can do more things, write, code, translate, understand a legal document, produce an image. In Intelligence as a Resource I developed the idea that when intelligence becomes an accessible resource, in the manner of electricity, the essential moves from the search for answers to the formulation of questions, an activity that engages what is most singular about us as human beings. To say that this tool makes us stupid is to mistake a redistribution of the power to act for a loss of capacity. And it is often those whose power is being redistributed who press the charge.
The presumption of competence
In cultural and educational work, I propose to draw a practical consequence from this and to reverse the burden of proof. Rather than starting from a supposed deficit, destroyed attention or lost culture, let us start from a presumption of competence: the young people we meet know things we do not, they have learned to navigate, to sort, to modulate their attention, skills I described in The Accelerated Eye with regard to accelerated viewing, and which offer footholds for transmission. Recent surveys of the engagement of 15-25-year-olds, which I took up in 15-25 year-olds have not deserted politics, point the same way on the civic front, where what is taken for desertion is a shift in the forms of engagement.
I am sometimes told that this stance would amount to giving up on high standards. Yet we only ask real work of someone we believe capable, and the presumption of competence is the very condition of any demand. It does not rule out vigilance about capture mechanisms, it makes that vigilance possible, since those mechanisms stem from interface design choices, and design choices can be debated, taught and regulated. As for the dumbing-down verdict, the Jyväskylä researchers, the authors of the great meta-analysis on short videos, Yann Leroux and Michel Serres, each with their own instruments, now give us good reason to see it for what it is, a very old reflex of power, and to consign it to the prop room.
