Multiple available measures indicate worsening mental health among US teenagers. Prominent researchers, commentators, and news sources have attributed this to effects of information and communication technologies (while not always being consistent on exactly which technologies or uses thereof). For example, John Burn-Murdoch at the Financial Times argues that the evidence “mounts” and he (or at least his headline writer) says that “evidence of the catastrophic effects of increased screen-time is now overwhelming”. I couldn’t help but be reminded of Andrew’s comments (e.g.) on how Daniel Kahneman once summarized the evidence about social priming in his book Thinking, Fast and Slow: “[D]isbelief is not an option. The results are not made up, nor are they statistical flukes. You have no choice but to accept that the major conclusions of these studies are true.”
Like the social priming literature, much of the evidence here is similarly weak, but mainly in different (perhaps more obvious?) ways. There is frequent use of plots of aggregate time series with a vertical line indicating when some technology was introduced (or maybe just became widely-enough used in some ad hoc sense). Much of the more quantitative evidence is cross-sectional analysis of surveys, with hopeless confounding and many forking paths.
Especially against the backdrop of the poor methodological quality of much of the headline-grabbing work in this area, there are a few studies that stand out as having research designs that may permit useful and causal inferences. These do indeed deserve our attention. One of these is the ambitiously-titled “Social media and mental health” by Luca Braghieri, Ro’ee Levy, and Alexey Makarin. Among other things, this paper was cited by the US Surgeon General’s advisory about social media and youth mental health.
Here “social media” is thefacebook (as Facebook was known until August 2006), a service for college students that had some familiar features of current social media (e.g., profiles, friending) but lacked many other familiar features (e.g., a feed of content, general photo sharing). The study cleverly links the rollout of thefacebook across college campuses in the US with data from a long running survey of college students (ACHA’s National College Health Assessment) that includes a number of questions related to mental health. One can then compare changes in survey respondents’ answers during the same period across schools where thefacebook is introduced at different times. Because thefacebook was rapidly adopted and initially only had within-school functionality, perhaps this study can address the challenging social spillovers ostensibly involved in effects of social media.
Staggered rollout and diff-in-diff
This is commonly called a differences-in-differences (diff-in-diff, DID) approach because in the simplest cases (with just two time periods) one is computing differences between units (those that get treated and those that don’t) in differences between time periods. Maybe staggered adoption (or staggered introduction or rollout) is a better term, as it describes the actual design (how units come to be treated), rather than a specific parametric analysis.
Diff-in-diff analyses are typically justified by assuming “parallel trends” — that the additive changes in the mean outcomes would have been the same across all groups defined by when they actually got treatment.
This is not an assumption about the design, though it could follow from one — such as the obviously very strong assumption that units are randomized to treatment timing — but rather directly about the outcomes. If the assumption is true for untransformed outcomes, it typically won’t be true for, say, log-transformed outcomes, or some dichotomization of the outcome. That is, we’ve assumed that the time-invariant unobservables enter additively (parallel trends). Paul Rosenbaum emphasizes this point when writing about these setups, describing them as uses of “non-equivalent controls” (consistent with a longer tradition, e.g., Cook & Campbell).
Consider the following different variations on the simple two-period case, where some units get treated in the second period:
Assume for a moment that traditional standard errors are tiny. In which of these situations can we most credibly say the treatment caused an increase in the outcomes?
From the perspective of a DID analysis, they basically all look the same, since we assume we can subtract off baseline differences. But, with Rosenbaum, I think it is reasonable to think that credibility is decreasing from left to right, or at least the left panel is the most credible. There we have a control group that pre-rollout looks quite similar, at least in the mean outcome, to the group that goes on to be treated. We are precisely not leaning on the double differencing — not as obviously leaning on the additivity assumption. On the other hand, if the baseline levels of the outcome are quite different, it is perhaps more of leap to assume that we can account for this by simply subtracting off this difference. If the groups already look different, why should they change so similarly? Or maybe there is some sense in which they are changing similarly, but perhaps they are changing similarly in, e.g., a multiplicative rather than additive way. Ending up with a treatment effect estimate on the same order as the baseline difference should perhaps be humbling.
How does this relate to Braghieri, Levy & Makarin’s study of thefacebook?
Strategic rollout of thefacebook
The rollout of thefacebook started with Harvard and then moved to other Ivy League and elite universities. It continued with other colleges and eventually became available to students at numerous colleges and community colleges.
This rollout was strategic in multiple ways. First, why not launch everywhere at once? There was some school-specific work to be done. But perhaps more importantly, the leading social network service (Friendster), had spent much of the prior year being overwhelmed by traffic to the point of being unusable. Facebook co-founder Dustin Moskowitz said, “We were really worried we would be another Friendster.”
Second, the rollout worked through existing hierarchies and competitive strategy. The idea that campus facebooks (physical directories with photos distributed to students) should be digital was in the air in the Ivy League in 2003, so competition was likely to emerge, especially after thefacebook’s early success. My understanding is that thefacebook prioritized launching wherever they got wind of possible competition. Later, as this became routinized and after infusion of cash from Peter Thiel and others, thefacebook was able to launch at many more schools.
Let’s look at the dates of the introduction of thefacebook used in this study:
Here the colors indicate the different semesters used to distinguish the four “expansion groups” in the study. There are so many schools with simultaneous launches, especially later on, that I’ve only plotted every 12th school with a larger point and its name. While there is a lot of within-semester variation in the rollout timing, unfortunately the authors cannot use that because of school-level privacy concerns from ACHA. So the comparisons are based on comparing subsets of these four groups.
Reliance on comparisons of students at elite universities and community colleges
Do these four groups seem importantly different? Certainly they are very different institutions with quite different mixes of students. They differ in more than age, gender, race, and being an international student, which many of the analyses use regression to adjust for. Do the differences among these groups of students matter for assessing effects of thefacebook on mental health?
As the authors note, there are baseline differences between them (Table A.2), including in the key mental health index. The first expansion group in particular looks quite different, with already higher levels of poor mental health. This baseline difference is not small — it is around the same size as the authors’ preferred estimate of treatment effects:
This plot compares the relative magnitude of the baseline differences (versus the last expansion group) to the estimated treatment effects (the authors’ preferred estimate of 0.085). The first-versus-fourth comparison in particular stands out. I don’t think this is post hoc data dredging on my part, knowing what we do about these institutions and this rollout: these are students we ex ante expect to be most different; these groups also differ on various characteristics besides the outcome. This comparison is particularly important because it should yield two semesters of data where one group has been treated and the other hasn’t, whereas, e.g., comparing groups 2 and 3 basically just gives you comparisons during fall 2004, during which there is also a bunch of measurement error in whether thefacebook has really rollout out yet or not. So much of the “clean” exposed vs. not yet comparisons rely on including these first and last groups.
It turns out that one needs both the first and the last (fourth) expansion groups in the analysis to find statistically significant estimates for effects on mental health. In Table A.13, the authors helpfully report their preferred analysis dropping one group at a time. Dropping either group 1 or 4 means the estimate does not reach conventional levels for statistical significance. Dropping group 1 lowers the point estimate to 0.059 (SE of 0.040), though my guess is that a Wu–Hausman-style analysis would retain the null that these two regressions estimate the same quantity (which the authors concurred on). (Here we’re all watching out for not presuming that the difference between stat. sig. and not is itself stat. sig.)
One way of putting this is that this study has to rely on comparisons between survey respondents at schools like Harvard and Duke, on the one hand, and a range of community colleges on the other — while maintaining the assumption that in the absence of thefacebook’s launch they would have the same additive changes in this mental health index over this period. Meanwhile, we know that the students at, e.g., Harvard and Duke have higher baseline levels of this index of poor mental health. This may reflect overall differences in baseline risks of mental illness, which then we would expect to continue to evolve in different ways (i.e., not necessarily in parallel, additively). We also can expect they were getting various other time-varying exposures, including greater adoption of other Internet services.
Summing up
I don’t find it implausible that thefacebook or present-day social media could affect mental health. But I am not particularly convinced that analyses discussed here provide strong evidence about the effects of thefacebook (or social media in general) on mental health. This is for the reasons I’ve given — they rely on pooling data from very different schools and students who substantially differ in the outcome already in 2000–2003 — and others that maybe I’ll return to.
However, this study represents a comparatively promising general approach to studying effects of social media, particularly in comparison to much of the broader literature. For example, by studying this rollout among dense groups of eventual adopters, it can account for spillovers of peers’ use in ways neglected in other studies.
I hope it is clear that I take this study seriously and think the authors have made some impressive efforts here. And my ability to offer some of these specific criticisms depends on the rich set of tables they have provided, even if I wish we got more plots of the raw trends broken out by expansion group and student demographics.
I also want to note there is another family of analyses in the paper (looking at students within the same schools who have been exposed to different numbers of semesters of thefacebook being present) that I haven’t addressed and which correspond to a somewhat different research design — which aims to avoid some of the threats to validity I’ve highlighted, though it has others. This is less typical research design, it is not featured prominently in the paper. Perhaps this will be worth returning to.
P.S. In response to a draft version of this post, Luca Braghieri, Ro’ee Levy, and Alexey Makarin noted that excluding the first expansion group could also lead to downward bias in estimation of average effects, as (a) some of their analysis suggests larger effects for students with demographic characteristics indicating higher baseline risk of mental illness and (b) if the effects are increasing with exposure duration (as some analyses suggest), which the first group gets more of. If the goal is estimating a particular, externally valid quantity, I could agree with this. But my concern is more over the internal validity of these causal inferences (really we would be happy with a credible estimate of the causal effects of pretty much any convenient subset of these schools). There if we think the first group has higher baseline risk, we should be more worried about the parallel trends assumption.
[This post is by Dean Eckles. Thanks to the authors (Luca Braghieri, Ro’ee Levy, and Alexey Makarin), Tom Cunningham, Andrey Fradkin, Solomon Messing, and Johan Ugander for their comments on a draft of this post. Thanks to Jonathan Roth for a comment that led me to edit “not [as obviously] leaning on the additivity assumption” above to clarify unit-level additivity assumptions may still be needed to justify diff-in-diff even when baseline means match. Because this post is about social media, I want to note that I have previously worked for Facebook and Twitter and received funding for research on COVID-19 and misinformation from Facebook/Meta. See my full disclosures here.]
Perhaps of interest:
https://causalinf.substack.com/p/did-facebook-hurt-our-mental-health
Thanks for this link. I thought my post was long! But this has interesting commentary that is relevant here, including about how it is possible to make the parallel trends assumption more plausible through placebo (negative control) outcomes, and what I quote below about TWFE giving a null result in the event study plot.
Interesting post, I’m curious what you make of Figure 2/the fully dynamic TWFE model as justification for the parallel trends assumption? I always found papers like this frustrating in grad school (and not just because I had to read all 30 pages). On the one hand clearly the authors have done their due diligence, nothing stands out as being obviously wrong, and the underlying hypothesis is believable. Yet on the other hand, when you start thinking about what’s actually going on and what’s actually being measured it all starts to seem so convoluted it’s hard to know what to take away from this study, particularly when it comes to external validity.
In an earlier version, I included Figure 2 as well. One of the goals with that figure is to highlight the absence of evidence for “pre-trends” (i.e. changes in differences between groups prior to the launch). I think it does a bit of that. Though it also highlights two things to me:
(1) There are pretty wide confidence intervals here, potentially allowing for a good deal of bias due to pretrends, but simply not stat. sig. Similar comments apply to Figure A.4 with this same kind of plot with just groups 1 and 4. Maybe it would have been nice to try the kind of sensitivity analysis of Rambachan & Roth here https://doi.org/10.1093/restud/rdad018.
(2) As Scott Cunningham (in the post Harlan links to above, https://causalinf.substack.com/p/did-facebook-hurt-our-mental-health) notes:
That is, one can think like a frequentist and wonder what would have happened if this study had been conducted before the whole “negative weights” diff-in-diff brouhaha. Would we have a published paper in the AER with that null event study? Etc.
As to your second more general point, that’s one reason I like directly matching on covariates and synthetic control by comparison.
Do you mean Suffolk County Community College on Long Island? There’s no mention of “Suffolk” after the title.
Yes, I am referring to SUNY Suffolk County Community College https://en.wikipedia.org/wiki/Suffolk_County_Community_College. I have edited the title to use the full name.
It has a launch date of May 26th, 2005, which is the last in this data set (while Harvard is the first). More generally, here it functions as an example of the different mix of institutions and students between expansion groups 1 and 4.
Thanks!
The biggest issue I can see with this particular article is the idea we can learn something from ~ 1 year or so of exposure to “thefacebook” a very early form of “social media”.
Cory Doctorow’s “Enshittification” cycle is absolutely a real thing, and the initial part of the cycle is to provide benefits to the users, to attract them. Therefore we’d really expect the early experience on FB to be beneficial to the users for the most part. And I remember certainly that it was. In 2009 ish I suddenly was connected to people I hadn’t talked to in years, and they were posting actual content about their own actual lives, and 99% of everything on the site was user created content. Within a few years it was “likes and reposts of news media” and today it’s 99% advertising and literally no-one I know is shown on my feed (even though when I go to their feed some of them are still posting things… because FB doesn’t show you stuff from your friends even if it’s there to be shown. When I realized that is when I completely ditched it for Mastodon).
My expectation given this dynamic, is that from 2004-2009 or so you had upticks in mental health caused by real-world connections to real people, but then overall mental health declined in the aftermath of the housing crash, the social media enshittification mid-cycle (where the big brands and businesses are the ones benefiting) and then the culture wars of the 2016 election and Trump and COVID etc. Somewhere in there FB died. It’s dead, and has been for a while now. The company still survives on its other products, the spy-advertising and the Instagram and etc, but FB itself is truly end-stage enshittification.
Any model that fails to account for that background knowledge is in my opinion quite limited.
Your first point is very much why I used “thefacebook” in the title and throughout here. More generally, I think it can be important to conceptualize communication technologies (which are quite plastic) in a way that allow us to go beyond treating them as black boxes. Some old arguments along those lines here: https://www.dropbox.com/s/cvtlsrau6a53rgy/Nass%20Mason.pdf?dl=0
With respect to “FB doesn’t show you stuff from your friends even if it’s there to be shown”, you might think about what you think of as the alternative or what other stuff you’re seeing (which I’m quite sure is mainly not actual paid ads). Have you read this paper? https://doi.org/10.1126/science.abp9364, which I also discussed here https://statmodeling.stat.columbia.edu/2023/07/27/new-research-on-social-media-during-the-2020-election-and-my-predictions/
I once counted I think 30 paid ads before I scrolled down to something posted by someone I know. I haven’t used FB more than 10 mins since December.
I’m enjoying mastodon though
Maybe there is a bunch of other content that you regard as being equivalent to “ads” (or is in some other third category besides “ads” and something posted by people you know). But I’m pretty sure there are rules that prevent several ads from appearing in a row in the feed, so I would be quite surprised if you saw 30 paid ads in a row. Hey I’d even be surprised by the top 2 positions both being ads.
I run a browser plugin called Social Fixer, so it’s conceivable it was confused at the time and making things worse. But even going back and turning off the plugin the vast majority of stuff is junk, though at the moment not pure ads, just essentially ads that people have “liked” onto the feed. They appear to have made some major changes to the format recently. In any case I’m extremely happy to no longer be using FB so they can get stuffed.
Checking now, of the first 25 posts in my Facebook feed, 15 are from friends and 10 are from groups that I either like or that FB thinks I would like. (Admittedly, its suggestions of groups are repetitive and not very good). Zero posts are ads. Perhaps it’s worse on mobile; perhaps advertisers have figured out that I don’t buy things; perhaps it’s luck. Anyway, I’ll be your friend, Daniel!
Raghu, it may very well be an interaction with a number of things. FB NEVER has ever been installed on any phone of mine. So this is always firefox, running in facebook container, running on a linux desktop machine with social fixer plugin.
Anyway a big part of what’s going on is of the 100 odd people I’m connected to, only about 4-5 of them post anything much anymore. I’m more than happy to be done with Facebook, but if you want to join the conversation on Mastodon I’m at @[email protected] I’ve had nothing but success over there with a bunch of scientists, linux enthusiasts, Julia programmers, infosec people, etc. Surprisingly low drama.
I mostly stopped using Facebook almost three years ago, although I haven’t completely given it up: I take a look a few times per month.
On the computer (not mobile) site at the moment:
First item is posted by a friend.
Next is “People You May Know”, suggesting other people I might add as friends.
Then an ad.
Then “Reels and Short Videos”
Then an ad.
Then three posts by friends.
Then an ad.
Then “Suggested For You”, which is not evidently an ad, in the sense that there’s no product they’re selling, but presumably they are hoping I will friend the site.
Then four posts by friends !!
Then two ads.
My situation is not quite as bad as Daniel reports, but it’s pretty bad.
That is with the default value for “Feeds”, which is “All.” When I change it to “Friends” the situation is better. Still sucky but better. But there doesn’t seem to be a way to change to “Friends” as the default, so every time I visit I have to click on Feeds and select Friends.
All in all it is a pretty crummy experience, which is why I don’t visit much anymore.
Correction! I forgot that I have an ad blocker (uBlock) on. Disabling it, 5 of the first 25 Facebook posts are ads.
Raghu, what percentage of the first 25 are actually content posted by your personal friends about their lives? A bike ride they took or a picture of their dog and a story about their walk in the park or an adventure in broken plumbing or some music they were listening to etc? That’s gone to nearly zero for me. What is posted by my friends is usually some outrage over a national news item or the like. And FB actually filters out stuff about my friends hikes or their kids etc (though most people aren’t even posting that stuff anymore)
Interesting; my (current) Facebook feed is mostly like what you describe as not existing — people I know, who are “posting actual content about their own actual lives”. It’s the reason I actually like Facebook (in contrast to Twitter, which I find awful in both form and content). Maybe your FB friends aren’t posting enough to populate your feed, or maybe the algorithm just doesn’t like you!
I wonder why Google -> Alphabet, Facebook -> Meta, and Twitter -> X since 2015?
I don’t have any idea, but it seems unusual for such large corporations to change their brand like that. So there is probably some reason common to all three.
Google and Facebook are too tied to a particular brand. They wanted to ambiguate their corporate holding company and decouple the stock from any particular product. The search engine is still Google and the main social media site is still Facebook.
Twitter is quite a different case, since it rebranded the actual product, is privately held, and has yet to spin up other products.
That is a valid difference. The similarity is that those are the three most popular companies whose primary revenue comes from people reading and typing things in *as the product*.
Seems like an interesting correlation to me.
“the three most popular companies whose primary revenue comes from people reading and typing things in *as the product*”
I don’t think Twitter/X is in the top three. Even not counting services with users highly concentrated in China, you have TikTok, LinkedIn, Telegram, and Snapchat. Depending on some variations in how these things are counted, Reddit and Pinterest are right there too.
https://en.wikipedia.org/wiki/List_of_social_platforms_with_at_least_100_million_active_users
I’m pressed for time and I haven’t read the paper, for which I apologize… but I immediately thought of something that I’m curious about: do they control for the 2004 election in some capacity? The geographical distribution of these universities seem to suggest that the last group have been differentially affected by the election and election results then the first group when it comes to the results. Do they have state dummies or the like?
Sidenote, Dean, I really like examples you use to illustrate the credibility issue of which diff-diffs work! Would it be ok for me to use this in a classroom setting, with attribution?
Yes, please feel free to use these. I recommend the corresponding section of the Rosenbaum book “Observation and Experiment” on this as well.
One of the biggest limitations here is that ACHA did not share the survey data with a school identifier, only with the expansion group and region I believe. I guess this was either for student privacy reasons or to prevent creating lists of “most depressed student bodies”.
Besides thinking about different post-election outcomes for Dem and Rep students, I would guess that the salience of political campaigns differs across these schools, with connections to differences in rates of working as volunteers and paid staff for national campaigns during and after college. Anyone know of research on that? I’m just drawing on personal experience and tangentially related studies like https://doi.org/10.1080/10584609.2015.1121941.
Thanks! And will try to look at your other recommendation too.
That’s an unfortunate limitation, although seems at least somewhat understandable.
Per your comment about the campaigns, yes! It’s the possibility of differentiated impacts overall due to politics that concerns me. I’m not familiar with any specific research on the topic, but post-2000 elections in the US *seem* to be ones that generated much more personal investment by the politically motivated, compared to before. So my guess is that this effect is something that would need be ruled out or at least accounted for.
As it stands, it’s interesting, but very limited evidence, not just because of this but also other things commentators have pointed out (e.g., the difference in the site in that time, the limited exposure)
Dean –
Interesting post.
I’m curious as to whether you’ve looked at the output from Jean Twenge regarding the impact of social media, particularly in teenage girls:
https://www.nytimes.com/2023/05/19/opinion/ezra-klein-podcast-jean-twenge.html
Every time I’ve looked closely at one of her papers, I haven’t found it very credible at all. I was thinking of this kind of work in the second paragraph of the post. Do you have any particular studies of hers in mind as being compelling, at least on first look?
Dean –
No.
I just heard her on a couple of podcasts, one more recently in Coleman Hughes’ pod. I didn’t listen that closely or follow up, but my general sense was she’s out over her skis somewhat – that she simplified the analytical task beyond what is reliable.
On the other hand, there’s something compelling about the conclusion she reaches, because I (like I think most people) have an instinctive reaction to kids spending so much time in front of screens and on social media more in particular.
What I thought most interesting is where she talked about sleep insufficiency as a mediating/moderating variable, and where she talked about the substitution of gaming for social media with boys relative to girls (which would explain less of a detrimental mental health impact on boys).
I believe Haidt relies on her work quite a bit in his anti-social media for kids advocacy, and I think he also tends to draw overly-broad conclusions at times.
Also, at the more popular media analysis level, I don’t know if you’ve seen this:
https://inews.co.uk/news/technology/dont-panic-about-social-media-harming-your-childs-mental-health-the-evidence-is-weak-2230571
Maybe it is worth following up on this more, particularly as “multiverse” or “specification curve” analysis plays a role in some of the back-and-forth here.
But, overall, without a better research design than just analysis of cross-sectional surveys, maybe this is all a distraction.
Problem is, we’re raising a 6-year old girl and so it may in fact be a distraction but it’s hard to not be concerned about a high level of risk.
Would like to read that response to Twenge and Haidt but I can’t get past the paywall.
This is a contentious field of research to say the least. And really difficult to study properly, because you can be absolutely certain that there is enormous effect heterogeneity – across individuals and usage patterns and their interaction.
As an example, a number of recent studies have looked at within-person change in time spent on social media and found no relation to within-person change i mental health problems. However, a major problem with these studies is that time spent is not necessarily the variable you are looking for, as the content of social media is algorithmically responsive to the observed attention of the individual.
Attentional and interpretational bias to certain stimuli is an important aspect of psychopathological states and their development. Take for instance the selective attention to and biased interpretation of body shapes in eating disorders. How are the algorithms of TikTok and Instagram likely to respond when an individual for one reason or another shows more attention to body shape and nutrition content? They will serve more of that kind of content, which I think is likely to often contribute to further development of eating disorder psychopathology, and increased attention to that kind of content, in a reciprocally reinforcing cycle. And remember, psychopathology is generally not discrete, but continuous from normal functioning. There are no bimodal distributions of psychopathological symptoms, they tend to left-skewed normal. People move across that continuum over time in developmental processes characterised by feedback loops.
The responsive nature of social media means that the amount of time used may be stable in an individual, and not strongly related to mental health, while changes in content may be reciprocally related to development of psychopathology. But studying this is really difficult, both methodologically, and without access to data from the social media companies themselves.
In my field (I’m a clinical child and adolescent psychologist) we see clear increases in the number of referrals for eating disorders (specifically) after the pandemic. I don’t *know* that algorithmically responsive image-based social media is a part of the cause, but I’m not at all convinced they are not involved in increasing the number of vulnerable individuals developing these disorders. So the question becomes one of where to put the burden of proof?
Erling: Good points about heterogeneity of what the treatment is. “The message is the medium.”
Another aspect of heterogeneity can be consumption vs. other users of social media, such as posting and directed interaction. See papers by Moira Burke and coauthors: https://doi.org/10.1111/jcc4.12162
As for the uptick with the pandemic, to me this also fits with an interactive role of social media use and other social interaction. As face-to-face interaction with peers is impeded, there is both more time for social media use and also perhaps more potential for distorted views of, e.g., descriptive norms on physical appearance etc.
Joshua, you should be able to access the Twenge critique at https://rdcu.be/b3C3o and the O&P response at https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7116236/ — but first read the original O&P paper at https://ora.ox.ac.uk/objects/uuid:5d844350-a359-47d3-b10c-4bfb93ce613b/download_file?file_format=application%2Fpdf&safe_filename=Orben-and-Przybylski-Nature-Human-Behaviour-AAM.pdf&type_of_work=Journal+article ;
for my critique of both Haidt and O&P see https://shoresofacademia.substack.com/p/haidt-social-media-in-o-and-p-paper ;
– David Stein
Nice study – thanks for the link. Yeah, I wouldn’t be surprised if social media also had positive influence on mental health for some or even many during the pandemic, by enabling social connections. Although the pre-social media internet managed the same, through all those forums and the like that were eaten up by facebook.
In any case, this really illustrates that “Is social media use or screen time harmful or nor?” is a way too general and broad research question. We should be asking questions about what sort of social media features, content or use patterns lead to what outcomes for which individuals – like Burke et al. do. I also find it striking that there are a number of studies looking at vulnerabilities within individuals related to poor outcomes of social media use, but I haven’t been able to find studies of social media product features that may have a negative impact. That kind of deflects the responsibility away from the profitable corporations.
Given the methodological challenges and potential confounding factors highlighted in your analysis of the study on the impact of thefacebook’s rollout on mental health trends, do you believe there are alternative research designs or strategies that could provide stronger evidence regarding the causal effects of social media on mental well-being among college students? Additionally, how might the findings of this study inform future research in the realm of technology’s influence on mental health?