Social Media and the Brain
Social media gets blamed for everything from silly comments to the spread of fake news and the erosion of tolerance and civility. Americans seem particularly concerned. A December 2022 Pew survey carried out in 19 countries found that U.S. respondents view the impact of social media on society and democracy more negatively than adults from other countries do.
Our concerns include the well-being of younger users. A report released on Tuesday found that three-quarters of American teenagers have seen online pornography by age 17, 38% specifically on social media platforms. In a 2021 newsletter, I described how Facebook's own research shows that Instagram undermines teen girls' mental health. The newest dangerous TikTok challenge is...well, I don't know, it's hard to keep track. Meanwhile, over a decade of research points to complex relationships between early social media use and problems ranging from peer rejection to depression.
In short, we have a lot to worry about. But since more than 70% of us now use social media, including more than 90% of teenagers, we need to understand exactly when and in what ways it's harmful, and to avoid exaggeration. In particular, we should avoid what psychologists call a reverse halo effect, or the tendency to assume that because something has a few negative characteristics, it must be, in general, a bad thing. Social media may be harmful in some respects, but it's not harmful for all people in all circumstances.
In this newsletter I'll be discussing new evidence that social media use changes adolescents' brains and makes them hypersensitive to feedback from peers. The study I'll be reviewing was widely covered in the news this week, including national outlets ranging from the New York Times to CNN, Fox, and Good Morning America. Although most of the reports touch on the study's limitations, the common theme is that it does tell us something about how social media influences developing brains.
Unfortunately, the study doesn't just have "limitations". It's one of the methodologically weakest pieces of research I've seen in a top-tier journal. And, its treatment in the news media illustrates why scientific findings can be intimidating. Complex methods and statistics, combined with state-of-the-art brain scan technology, gives the study a veneer of credibility that it doesn't deserve.
If I seem excessively cranky, it's because what's at stake here are beliefs about potentially permanent changes to young brains. If the study is fatally flawed, parents, educators, and others might be needlessly alarmed (or encouraged, if one happens to view the findings positively).
I'll describe the study first, then call out a few of its more egregious weaknesses. At the end I'll touch on the broader question of what we know so far about the impact of social media on adolescent neurodevelopment.
The new study
The study I've been alluding to was published in the prominent journal JAMA Pediatrics on January 3. Maria Maza and colleagues at the University of North Carolina looked at associations between social media checking among adolescents and their brain development across a three-year period.
At the beginning of the study, 169 6th and 7th graders were asked how many times per day they use three popular social media platforms (Facebook, Instagram, and Snapchat). At that time, and then twice over the next two years, the adolescents also completed a Social Incentive Delay (SID) task while receiving an fMRI.
The SID task is a simple computer game in which players press a button as quickly as possible once a target appears on the screen. The target only remains on the screen for a short time (e.g., a third of a second). If players press the button before the target disappears, they're rewarded with a cartoon image of a smiling face. If they press the button too slowly, they see a frowning face. These faces are the only social feedback provided.
One more detail of the SID task is crucial: Before the target appears, players either see a plus sign (+) indicating that a smiling face will appear if they respond quickly enough, or a minus sign (–) indicating that a frowning face will follow a slow response. The purpose of the plus and minus signs is to create a sense of anticipation during the few seconds before the target appears.
The researchers wanted to know what happens in adolescents' brains during those few seconds that social feedback is anticipated. To find out, adolescents completed the SID task while hooked up to a functional magnetic resonance imaging (fMRI) machine. An fMRI shows, in real time, which areas of the brain become more or less active as a person engages in some thought process or behavior. In this study, the fMRI showed which areas were most active while the adolescents were anticipating social feedback.
The main findings emerged from a comparison between "habitual checkers" (adolescents who checked their social media 15 or more times per day) and the rest of the sample (those who checked their social media 1 to 14 times per day). Across the three years of the study, sensitivity to social feedback increased among the habitual checkers while decreasing among everyone else.
"Sensitivity" is the researchers' umbrella term for neural activity in several areas of the brain (amygdala, ventral striatum, etc.) that are linked to experiences such as receiving rewards, taking risks, deciding what's important in one's immediate surroundings, and regulating one's emotions. Increased sensitivity among the habitual checkers was somewhat consistent across brain areas. Thus, the researchers concluded that frequent social media use leads to hypersensitivity to social feedback.
What does "hypersensitivity" mean exactly? Hard to say – the researchers don't explain it – but, based on what other experts have written, you might think of it as neural activity that would make you highly motivated to seek social feedback, and highly responsive to whether that feedback is positive or negative. Is it a good thing then? Also hard to say. In interviews with national news media, the researchers equivocate, claiming that hypersensitivity might be desirable (if it means that adolescents are more motivated to connect with others) but it might be undesirable (if it makes teens oversensitive to negative feedback). In the study itself, hypersensitivity is implicitly treated as a negative, "divergent" pattern of development associated with excessive checking of social media. This is the message most likely to be gleaned by pediatricians and others who work with adolescents.
Evaluation of study
The new study addresses an important question – does social media use change young peoples' brains – but I don't think the results tell us anything, in spite of the impressive fMRI technology and some high-octane statistical analyses.
One of the study's most serious flaws is that it reveals almost nothing about adolescents' actual social media use.
The researchers asked participants how often they used Facebook, Instagram, and Snapchat. No explanation was given for exclusive focus on these three platforms, but we know they're not the only ones adolescents use. For example, during the time this study was conducted, more teenagers were using TikTok than each of those three platforms.
A further problem is that adolescents were only asked at the outset of the study how often they check their social media. One would expect that across a three-year period, at least some adolescents would be on social media more or less frequently than before.
The relationship between social media checking and sensitivity to social feedback is probably complex and multi-directional. Some experts have suggested that teens who are more sensitive to social feedback check their social media more often, and the feedback they receive further increases their sensitivity. The researchers did not evaluate hypotheses like this. Instead, they assumed a simplistic causal model in which social media use at an earlier time might influence sensitivity later on.
Even if we assumed that a one-time measurement of social media usage is good enough, the way responses were coded pretty much guarantees an inaccurate read. Here the researchers took a flagrantly post hoc approach, adjusting their coding scheme repeatedly until they found one that yielded significant results.
I want to step into the weeds for a moment, because this a good illustration of the intimidation factor I mentioned earlier. The coding of social media use in this study is convoluted enough to seem technical and "sciencey" – none of the news reports I saw raised questions about it – but once you look more closely, the flaws are obvious.
Here's how the researchers initially introduce their coding scheme:
"[P]articipants were asked how many times per day they checked each platform, with answers grouped into 8 numerical score categories (1, <1 time per day; 2, 1 time per day; 3, 2-3 times per day; 4, 4-5 times per day; 5, 6-10 times per day; 6, 11-15 times per day; 7, 16-20 times per day; 8, >20 times per day)."
In other words, an 8-point scale was created to describe how many times per day each adolescent checked their social media. No rationale is provided for this scale, but, in any case, most researchers would stop here and immediately begin data analysis. Instead, the researchers proceeded to recode their initial coding:
"We recoded participants’ scores to create an ordinal scale that captured social media checking frequency across a meaningful distribution that could be assessed quantitatively. A score of 1 was recoded to 0 and a score of 2 was recoded to 1. Scores between 3 and 7 were recoded to the average of the range of number of times checked; for example, if participants...checked Facebook between 11 and 15 times per day, then their score was recoded to...13 times checked. Reported scores of 8 (ie, checked >20 times per day) were recoded to 20 times checked. For each participant, the recoded checking behaviors on the 3 social media platforms were summed to create a total social media checking score that ranged from 0 to 54..."
I'm overwhelmed by how much is wrong with this passage, so I'll just focus on a few of the key problems.
Regarding the first sentence, it's untrue that the original scale could not be assessed quantitatively, and that the recoded scale would be better, or more meaningful, or whatever.
The phrase "ordinal scale" simply refers to a scale arranged in order from highest to lowest, where the difference between adjacent points may not be consistent. The researchers started with an 8-point ordinal scale, where "8" means checking social media 20 or more times per day, "7" means checking 16 to 20 times per day, and so on. They then shifted to another ordinal scale, both of the type routinely used by social scientists. In short, a seemingly arbitrary choice of one coding scheme was replaced by a seemingly arbitrary choice of another one. (Stats people: If the raw score distribution was problematic – e.g., deviated quite a bit from normality – it wouldn't make sense to create ordinal divisions that fail to fix the problem, if an ordinal approach were even justifiable to begin with.)
Regarding the last sentence quoted above, after recoding the data to create a new ordinal scale, the researchers then summed responses across platforms to create a total score ranging from 0 to 54. The cherry on top of this dreadful cake is their final step – once again arbitrarily chosen – of using that total score to divide participants into three groups: Habitual users (15 or more checks per day), moderate users (1 to 14 checks per day), and nonhabitual users (less than 1 check per day).
The main findings of the study emerged from a comparison between the habitual group versus the other two groups combined. If this way of carving up the sample was most appropriate, the researchers could've started with it. Instead, they introduced it at the end. It looks arbitrary, and it doesn't make a lot of sense. For instance, how can you assign someone who checks their social media once or twice per day to the same group as someone who checks 13 to 14 times per day? In what sense are those people both "moderate" users?
Since it seems unlikely that the researchers' choices were truly arbitrary, the only thing I can conclude is that they chose the coding scheme that yielded the most favorable results.
Fine, you might say, the study still hints at something important. It hints that adolescents who check their social media a lot develop hypersensitivity to social feedback, even if we have no idea what the cut-off might be for "a lot".
I'm not persuaded by this line of reasoning, because the potential for measurement and coding error is so great, and because the design of the study is too simple to rule out a range of alternative interpretations. To take just one example, a third variable such as social skills could be responsible for the longitudinal trend that emerged. Less socially skilled adolescents might be the ones who use social media the most and, as they get older, thes one who become most sensitive to social feedback.
Should we worry?
In my view, this study provides no evidence that social media harms (or benefits) adolescents' brains.
As for other studies, although it's difficult to prove the negative, I don't think there's clear evidence yet that social media per se affects adolescent neurodevelopment. The main reason is that existing studies don't uniquely implicate social media processes. Rather, they tell us about the impact of experiences that might occur on social media and during face-to-face interactions. For example, victims of cyberbullying tend to have impaired sleep, and impaired sleep has been linked to micro-structural brain changes that impair neural connections. Thus, one might expect cyberbullying to affect victims' brains, but there's no reason to expect the effects of face-to-face bullying to be any different. Peer rejection, exclusion, and other forms of negative social feedback also seem to impact adolescents' brains, but what hasn't been shown is that the effects are distinctive to social media interactions.
I wouldn't view findings like this as good news, but rather as reflecting an absence of bad news. At the moment we shouldn't worry that social media use per se impairs the development of young brains. We should be concerned about adolescents who use social media excessively, or have persistently negative experiences on social media, but our concerns should be about their mental health and social functioning rather than their neurodevelopment.
Thanks for reading! Feel free to share a link on your social media account – I promise you won't damage anyone's brain.