I'd love to see debate between Jon Haidt and his detractors on this topic--like that 2015 (?) SPSP when he went head-to-head with Kurt Gray--because I find his claims to be pretty convincing and include references to causal studies, and I see critiques like Professor Przybylski's above (caveat: I only watched the first video) tend to make a lot of "correlation!=causation" without addressing the causal evidence that exists.
Specifically, in your discussion there was an assumption there is no experimental evidence for reduced well-being from social media. But Haidt, Rausch, & Twenge's ongoing review [1] currently lists 38 "experiments using random assignment that indicate a causal effect on mental health outcomes". And perhaps no one has specifically run this on children, or maybe the evidence is mixed (Haidt et al's doc also lists another 9 experimental studies that found no effect) but I get the feeling of two groups talking past each other when one side lists 47 mixed experimental findings and the other side seems to be saying there is no evidence at all. I haven't really taken the time to assess those 47 mixed experimental findings--maybe none are relevant or meaningful--but it seems like they need to be grappled with to have a serious debate on this topic?
After a lengthy discussion where Przybylski describes why he thinks Haidt doesn't have the evidence he needs to support his own claims, he then asserts without describing any evidence at all that "abstinence doesn't work". I don't think the analogy to sex abstinence education really carries over any lessons about, say, a school mandating that its students keep cellphones out of the classroom. Surely evidence is required to assert a null effect equally as much as when asserting an effect?
Increasingly, I think I just need to write a deeper white paper on what the data actually say. Sure, every technology, social media and smartphones included, harm some people some of the time in some use cases. Yet, the public causal evidence for it is quite weak. The studies have issues with sampling (sometimes even snowball sampling teens who display "problematic social media use" which leads to anxious teens bringing even more anxious teens into a study, which amounts to an overrepresentation of anxious teens in these studies), measurement (invalid measures, like self-reported time spent on social media each day), analytic strategies (p-hacking), questionable research practices (weird exclusion criteria, differing covariates with no real justification of why they differ across analyses, no pre-registration of sampling or analytic approaches or code, no data sharing, etc), inadequate attention given to the reverse causal direction (we know that depressed people tend to overestimate in their self-reports of how much time they spend doing things, including social media use... and, we also know that when someone becomes more depressed it increases their likelihood of using social media more than using social media increases depression), and more.
Anything involving social networks is complicated, and when those social networks are run by privately held companies that are doing their best to withhold their data from the public eye, it's even tougher. Jon is a very compelling speaker, and he knows how to piece together data to support an excellent, very believable story. Yet, the story he's telling is overly simplistic. It's hard to argue that any technology doesn't cause harm for some people some of the time. I've spent the last 5+ years fighting to reduce harms on social media, and I worry that his compelling story.
I've got a few more interviews scheduled with some other researchers who work in this space, and I'll push them to get more into some of these details. Chris Ferguson is next up in my queue, and he just published this meta-analysis that goes into some of these issues here: https://christopherjferguson.com/Social%20Media%20Meta.pdf. Another good review that's now a couple years old is this one by Matt Vuorre, Amy Orben, and Andy (https://journals.sagepub.com/doi/abs/10.1177/2167702621994549).
We go into more detail into the abstinence bit in part 2 of our conversation... but, there is good evidence that coddling kids and not preparing them for things they will encounter at some point in their lives usually leads to bad outcomes. Jon wrote an entire book about that.
I'd love to see another debate like that one on morality with Kurt a few years back, too. UVa hosted a good exchange between he and Candace Odgers a couple weeks back. Candace doesn't have the same dramatic flair that Kurt does, but the conversation was excellent. You can see that here: https://tyde.virginia.edu/event/haidt-odgers/
I'd love to see debate between Jon Haidt and his detractors on this topic--like that 2015 (?) SPSP when he went head-to-head with Kurt Gray--because I find his claims to be pretty convincing and include references to causal studies, and I see critiques like Professor Przybylski's above (caveat: I only watched the first video) tend to make a lot of "correlation!=causation" without addressing the causal evidence that exists.
Specifically, in your discussion there was an assumption there is no experimental evidence for reduced well-being from social media. But Haidt, Rausch, & Twenge's ongoing review [1] currently lists 38 "experiments using random assignment that indicate a causal effect on mental health outcomes". And perhaps no one has specifically run this on children, or maybe the evidence is mixed (Haidt et al's doc also lists another 9 experimental studies that found no effect) but I get the feeling of two groups talking past each other when one side lists 47 mixed experimental findings and the other side seems to be saying there is no evidence at all. I haven't really taken the time to assess those 47 mixed experimental findings--maybe none are relevant or meaningful--but it seems like they need to be grappled with to have a serious debate on this topic?
After a lengthy discussion where Przybylski describes why he thinks Haidt doesn't have the evidence he needs to support his own claims, he then asserts without describing any evidence at all that "abstinence doesn't work". I don't think the analogy to sex abstinence education really carries over any lessons about, say, a school mandating that its students keep cellphones out of the classroom. Surely evidence is required to assert a null effect equally as much as when asserting an effect?
[1] https://docs.google.com/document/d/1w-HOfseF2wF9YIpXwUUtP65-olnkPyWcgF5BiAtBEy0/edit?tab=t.0#heading=h.npba6tvfpq0g
Hey Ben! Thanks for the comment and question.
Increasingly, I think I just need to write a deeper white paper on what the data actually say. Sure, every technology, social media and smartphones included, harm some people some of the time in some use cases. Yet, the public causal evidence for it is quite weak. The studies have issues with sampling (sometimes even snowball sampling teens who display "problematic social media use" which leads to anxious teens bringing even more anxious teens into a study, which amounts to an overrepresentation of anxious teens in these studies), measurement (invalid measures, like self-reported time spent on social media each day), analytic strategies (p-hacking), questionable research practices (weird exclusion criteria, differing covariates with no real justification of why they differ across analyses, no pre-registration of sampling or analytic approaches or code, no data sharing, etc), inadequate attention given to the reverse causal direction (we know that depressed people tend to overestimate in their self-reports of how much time they spend doing things, including social media use... and, we also know that when someone becomes more depressed it increases their likelihood of using social media more than using social media increases depression), and more.
Anything involving social networks is complicated, and when those social networks are run by privately held companies that are doing their best to withhold their data from the public eye, it's even tougher. Jon is a very compelling speaker, and he knows how to piece together data to support an excellent, very believable story. Yet, the story he's telling is overly simplistic. It's hard to argue that any technology doesn't cause harm for some people some of the time. I've spent the last 5+ years fighting to reduce harms on social media, and I worry that his compelling story.
I've got a few more interviews scheduled with some other researchers who work in this space, and I'll push them to get more into some of these details. Chris Ferguson is next up in my queue, and he just published this meta-analysis that goes into some of these issues here: https://christopherjferguson.com/Social%20Media%20Meta.pdf. Another good review that's now a couple years old is this one by Matt Vuorre, Amy Orben, and Andy (https://journals.sagepub.com/doi/abs/10.1177/2167702621994549).
We go into more detail into the abstinence bit in part 2 of our conversation... but, there is good evidence that coddling kids and not preparing them for things they will encounter at some point in their lives usually leads to bad outcomes. Jon wrote an entire book about that.
I'd love to see another debate like that one on morality with Kurt a few years back, too. UVa hosted a good exchange between he and Candace Odgers a couple weeks back. Candace doesn't have the same dramatic flair that Kurt does, but the conversation was excellent. You can see that here: https://tyde.virginia.edu/event/haidt-odgers/