A global review reveals links between screen use and various child health outcomes

This is a methodologically very sound article. It is a review of reviews that brings together 23 meta-analyses and compiles data on nearly three million children from 64 countries; the sample is therefore extremely large. It has been pre-registered, assesses the quality of the studies it includes and – a rare and methodologically sound practice – excludes studies that appeared in multiple meta-analyses, thereby preventing the results from being inflated. Furthermore, the authors also report on data that did not prove to be statistically significant, which lends transparency and scientific rigour to the study as a whole.

However, the main findings confirm what research has been showing for years and which puts the current media and social moral panic surrounding the relationship between young people and screens into perspective: there is an association between screen time and poorer indicators of mental health and academic performance, but it is very small (around 1 per cent of the differences between one group of adolescents and another). This is not an insignificant finding, but it in no way supports the mantra that screens are the main cause of adolescent emotional distress. As we have pointed out in previous articles, screen use alone does not explain why a teenager is emotionally distressed or performs worse academically.

Other caveats to bear in mind when reading the results are as follows: firstly, some striking findings, such as emotional problems or difficulties with peers, are based on very few studies and show significant variability between them. Their confidence intervals range from an almost negligible effect to a moderate one, so these are the least reliable estimates in the article, and they are the ones most likely to make headlines in the media, as is often the case. Secondly, the authors compare the magnitude of these effects with that of socio-economic status, and this comparison must be treated with caution, as studies on screen use typically do not use socio-economic status as a variable to adjust their results. Moreover, as argued by some in contemporary educational sociology, we might consider that some of what is attributed to screen use may, in reality, be social inequality expressed through these devices.

Perhaps the most novel aspect of the study lies in the distinctions it draws between different types of screen use. For example, the association between such use and psychological distress in children is stronger with mobile phones than with television or video games, and stronger with depressive symptoms than with social anxiety. Furthermore, this association is more pronounced in adolescents than in younger children. All of this delves into another issue that has been highlighted by the latest research on the subject: it matters less how many hours devices are used for and more which device is used, for what purpose, with whom and at what stage of development. Recommendations based solely on total hours lump together very different types of use, as playing online with friends or doing homework is not the same as passively scrolling for hours through a feed designed to hold one’s attention.

Another limitation worth noting is that, as is often the case in such studies, most of them are cross-sectional and do not allow for the establishment of causality, only association.

Finally, it is worth highlighting another issue commonly raised in studies of a more sociological nature than the one at hand—although this study clearly demonstrates it: the most significant factor in the association between children and screen time is the time spent scrolling. In other words, the problem is not screen use itself, but rather what I fail to do whilst immersed in it, such as sleeping, engaging in physical activity, or socialising with family and friends.

In the field of education and technology use – which is my area of expertise – I have been able to analyse the results in detail. Thus, when there is talk of a negative correlation [between screen use] and academic performance, this should be interpreted with considerable caution. The findings are highly varied, making it difficult to establish a causal link. Furthermore, in terms of educational research, this study suffers from the same shortcoming as other primary studies: the purpose of use, the content or the level of interaction with screens are aspects that are not usually taken into account (the study makes a slight distinction between types of screens, but not what is actually done with them). Other studies that have taken these factors into account have found that no such link exists. It is therefore important to draw a clear distinction between correlation and causation.

By this I mean to emphasise that whilst these findings are significant, they must be treated with caution (as the authors themselves note in their discussion of limitations) and that further analysis of contextual factors not previously considered is required.

This is a high-quality study that brings together evidence from nearly 3 million young people across the globe. It shows a relationship between overall screen time and some developmental and mental health outcomes.

It cannot draw robust conclusions about specific types of screen use, like social media use or gaming, as there was not enough data to study this. We should therefore be cautious in interpreting these results in the context of young people’s current screen use habits.

Most of the data contributing to these conclusions comes from cross-sectional studies, where screen use and the developmental or mental health outcome are measured at the same time. This means that it cannot say much about whether higher screen use leads to poorer outcomes.

Research is urgently needed to understand the role of different types of screen use in the development of young people’s developmental and mental health outcomes over time to support policy making in this area.

In my view, this article once again highlights a recurring problem in recent educational research and the direction it has taken as a result of two factors that have converged over time. On the one hand, the influence of certain perspectives on evidence and causality that are closely linked to experimental and quasi-experimental designs; and, on the other, the rise of academic capitalism.

We have 23 meta-analyses, involving more than 2.8 million children from 64 countries, and a vast array of statistical analyses, all to conclude that there is an association between increased screen time and certain outcomes relating to development, academic performance or mental health.

However, the question that arises for me is: how much does all this actually help us to understand the phenomenon under study and, above all, to make decisions based on that understanding? Very little, because this type of research can describe certain associations with great precision, but describing an association is not the same as understanding the phenomenon that causes it.

Knowing that two things are associated tells us relatively little about why they are, through what mechanisms, or under what conditions. And the problem, in this specific research, is exacerbated when we look at what we mean by ‘screen time’. This category encompasses radically different experiences: passively watching videos, playing games, creating content, talking to family members, socialising with friends, using social media, or engaging in educational activities. The authors themselves acknowledge that much of the research conflates qualitatively different activities and that fundamental dimensions such as content, context or the specific characteristics of usage are left out of their analyses.

The problem, therefore, runs deeper than a mere methodological limitation. It lies in the way we define the research problem. When we ask ourselves what effects ‘screen time’ has, we have already made an important theoretical decision: to reduce vastly different technological experiences to a single quantifiable variable. What we choose to measure conditions the measurement itself and, even before that, our own construction of the phenomenon. We have constructed ‘screens’ and ‘time’ as homogeneous entities and as categories relevant to explaining the phenomenon. We design our research on the basis of this theoretical construction.

We could spend our lives increasing sample sizes, accumulating studies and greatly refining our statistical analyses. But none of these things will resolve a conceptual problem in the construction of the object of study. It will not help us to understand the phenomenon and, therefore, nor will it help us to make decisions based on that understanding.
This makes a finding within the article itself particularly significant: extraordinarily high levels of heterogeneity in many of the results.

Added to this is the problem of causality. Most of the findings are correlations (in education it is very difficult to establish causality, which is why experimental and quasi-experimental methods have particular limitations in this field) and the authors themselves expressly caution that it cannot be concluded that screen time causes the observed outcomes.

That is why I return to the previous question: what specific decision can we make based on this information? Should we reduce ‘screen time’? Which types of screen time? To achieve what? At what ages? In what contexts? Replacing which activity? If certain forms of use may be harmful and others beneficial (something the article itself acknowledges), knowing that ‘more screen time’ is statistically associated with certain outcomes tells us very little indeed. In fact, the authors themselves raise the possibility that beneficial and harmful experiences may even be cancelling each other out when they are all lumped together under the category of ‘screen time’.

There is, moreover, something about the article that strikes me as particularly striking. It acknowledges virtually all of these limitations and yet, when it comes to establishing the implications of its findings, it seems to act as if these had far fewer consequences than it has just explicitly acknowledged. It acknowledges that it is aggregating very different phenomena, that key variables for understanding screen use are missing, and that the evidence is predominantly correlational; yet it ends up presenting screen use as the factor to be modified and arguing that the results can guide policy on children’s digital use.

There is, moreover, an interesting paradox here. When it comes to explaining the associations it finds, the article itself has to resort to possible mechanisms that its design does not allow it to study directly: shifts in sleep, physical activity or face-to-face interaction; processes of social comparison; platform design features; and so on. In other words, when we want to move from describing what is associated with what to trying to understand why, we need precisely what this type of research does not study.

Perhaps the problem is that we need less research devoted to continually asking how much ‘screen time’ is associated with a particular outcome, and much more research aimed at understanding the ‘whys’: what children and adolescents actually do with technology, what meaning it holds for them, what they use it for, with whom, in what contexts, and through what processes certain uses end up being beneficial or harmful.

And that probably also calls for a different kind of research. If we want to understand mechanisms, contexts, experiences and meanings, we will need far more qualitative (or mixed-methods) research, rather than simply continuing to increase sample sizes and add layers of statistical sophistication. The problem is that such research is far less eye-catching when it comes to publication in high-impact journals (and here it is worth criticising the scientific publication system). A few dozen in-depth case studies make far less of an impression than ‘2.8 million people under the age of 64 from 64 countries’ and are less suited to a scientific culture in which large samples, meta-analyses and statistical sophistication are the most highly valued elements.

All this leads me to something that has concerned me for quite some time: the relationship between certain social discourses and the production of scientific knowledge works both ways. Science produces findings that fuel and legitimise certain discourses on ‘screens’, but those very same discourses also shape what we consider to be a problem worthy of investigation and what questions we ask about it.

If we have socially constructed ‘screens’ and ‘screen time’ as the problem, it becomes much easier for research to begin by asking how much time children spend in front of them and what the consequences are, rather than asking what they are actually doing with the technology, for what purpose, how and why.

And so we come full circle: a complex, heterogeneous and largely correlational finding enters the public debate as ‘screens are harmful to children’. That narrative in turn fuels new research questions formulated in the same terms. And here, journalism also has an important responsibility: to explain these issues rigorously, rather than chasing the easy headline.

The problem, of course, is not studying the possible negative effects of technology. That needs to be done. The problem is doing so without attempting to understand the phenomenon, limiting ourselves to describing associations and basing our work on highly questionable theoretical constructs right from the outset in defining the research problem.

Perhaps we do not simply need larger studies along the same lines, but research that seeks to understand and begins by asking different questions.