Social Media as a Cause of Mental Health Problems

Longstanding concerns exist about excessive social media use causing mental health problems. The best evidence for this view comes from longitudinal studies showing that earlier social media use leads to later mental health problems, and from experimental studies in which reductions in social media use lead to improved wellbeing. However, recent articles have criticised the strength of this evidence^1,2,3. Given the pressing importance of understanding whether social media is psychologically damaging, these criticisms deserve consideration.

Vast numbers of people use social media daily, if not hourly. If such usage has the potential for impacting mental health negatively, then this is a major concern for already overstretched clinical services. If it does not have this potential, then effort is better focused elsewhere. Given this, examination of causal claims about social media use is to be welcomed. However, this examination must understand not only the components of valid science, such as study design and statistical tools, but also what such research can illuminate for clinical purposes. It may be that these critiques of the evidence have merit scientifically, but are still wide of the mark clinically.

Leaving aside cross-sectional investigations showing correlations between social media use and poor mental health, two sets of studies are causally important. Some studies document temporal relationships between social media use and mental health. They show high level social media use at time one is associated with worse mental health at time two. Other studies experimentally manipulate social media use. These show that groups who reduce their social media use subsequently display better mental health than groups not reducing usage. These data cannot easily be ignored, but their evidential strength in support of a causal claim can be analysed.

At least three recent papers^1,2,3 suggest this causal evidence is weaker than people may imagine. They suggest that: the observed relationship is very small; the measures of social media usage are blunt; and other factors may better explain these relationships. However, when these counterclaims are evaluated, then the strength of their arguments is not as clear as the headline messages. All of these criticisms have some scientific merit, but also some flaws, and all make common misassumptions about what scientific research can do in terms of illuminating clinical phenomena; perhaps it is a rather common case of scientific overreach.

The first criticism is that effect sizes in longitudinal studies are very small, meaning the relationship is not important^1,3: “Surveys that track individuals over time suggest that…social media use only predicts a small change in well-being over time.”¹. Now, while some studies do find small relationships, others find larger-sized effects. Which studies are accepted is a matter of judgement, not helped overly by using averaging or multiple analytic techniques³, which provide no answers to the questions raised. However, this is statistical quibbling of little relevance in this context. The size of a correlation between a causal event and clinical outcome often has no relationship to the importance of that association. For example, traumatic events do not, often, last for long (it may not feel like this is the case for those experiencing them, but that is an important point for another argument). Given this, the relationship between traumatic time and its consequences is not that strong, which has been shown empirically: “…for externalizing problems…lesser proportion of exposure was associated with more severe symptoms. For internalizing problems, neither age of onset nor proportion of exposure were predictive of symptoms severity.”⁴ Despite this, nobody seriously claims that traumatic events are not causal for mental health difficulties. The same could be true for screen time and mental health.

A short meaningful event produces a cascade of associations, each contributing to a nexus between that event and subsequent mental health problems. Trauma may have its direct impacts on brain structure and function, but it also makes people behave differently, which can generate unique problems and amplify the initial issues – but without the initial trauma, the cascade would not occur. The same may be true of small amounts of social media use; this will depend on the effects for the individual, about which large-scale population studies are silent.

This leads to the consideration of time-based metrics for social media use. It may be that: “…time alone is a poor metric to gauge effects.”¹ However, this actually undermines the negative conclusions drawn by the same papers^1,2,3 that suggest relationships between social media and mental health problems are: “…trivial and differ little from the impact of other meaningless correlations…(e.g., Orben & Przybylski, 2019).”² At best, this consideration reflects negatively about the measure, but not the construct. Screentime may be a poor metric, but it still cannot be said: “…nobody would argue we should study the causes of obesity by investigating "food time"¹. This analogy is misleading in clinical terms. If somebody spent all their time eating (high food time), they would get fat and unwell; conversely, if they spent none of their time eating (low food time), they would get ill and die. There is a clinical relationship between time and outcome, but not one easily captured by population-level statistics using samples containing many without mental health problems. Indeed, that there is any relationship between screentime and poor wellbeing across such populations indicates that there is something important here.

The third criticism suggests that, when relevant variables are controlled, relationships between social media use and poor mental health disappear. One study noted that associations between social media use at nine years and mental health at 13 years disappeared when emotional problems, language delays, family conflict, happiness, and popularity, were controlled for, concluding: “…these results do not support the belief that social media use, at least in regard to general use or time spent on social media, is associated with negative outcomes for youth.”³. This is like saying: once you take away all the mental health variables that social media use could affect, then social media does not affect mental health. Imagine saying: traumatic events have no effect on functioning, once you control for flashbacks, nightmares, avoidance, and physiological changes. Moreover, these behaviours may be part of the cascade of influences forming the causal nexus between screentime and mental health problems. Their removal from consideration will, as a matter of course, remove the statistical association, but will be clinically misleading.

Thus, while the scientific reasoning of the critical studies may have some validity, the logic and applicability to clinical phenomena are questionable. Science has its uses, but the important thing is knowing the limits of generality for the claims. Large-scale studies can guide clinical knowledge, and produce useful rules-of-thumb, but they do not work at a specific level for all individuals.