Teens and Tech: Moral Panic or Moral Emergency?
MAY 29, 2019
Has social media compromised the mental health of a generation? As we wrote in a prior newsletter, the question is far from settled, despite a proliferation of headlines and popular books. Some have argued we’re in the midst a public health crisis; others, a moral panic. The research itself remains mixed, and, like many recent scientific controversies, it is deeply entangled with questions of reproducibility and research quality. A recent PNAS paper by Amy Orben et al. represents a new volley in the ongoing debate, arguing that “social media effects are nuanced, small at best, reciprocal over time, gender specific, and contingent on analytic methods.”
To put things in context: the present wave of concern about screens and teens began in earnest with the publication of Jean Twenge’s 2017 book iGen, which argued that digital technology use had contributed to delayed development and worsening mental health among kids born between 1995 and 2012. Twenge’s research was subsequently picked up by Jonathan Haidt and Greg Lukianoff in The Coddling of the American Mind, which tied Twenge’s work to broader trends on US college campuses. Both claims -- that teen mental health is worsening, and that digital technology is a culprit -- have been the subject of much debate, prompting Haidt & Twenge to create semi-open-source literature reviews on each topic.
Orben and her colleague Andrew Przybylski helped to spur this response back in January, when they re-analyzed the data from one of Twenge’s papers and published the results in Nature Human Behavior. The problem, they explained at the time, is that massive datasets of the kind used in Twenge’s study offer too much analytical flexibility to the researcher, making it “almost impossible not to find statistically significant effects.” To get around this, Orben and Przybylski made use of a novel statistical technique known as specification curve analysis. This allowed them to run tens of thousands of theoretically defensible analyses on the same datasets, mitigating the hindsight bias that can affect any single post-hoc analysis. The result? Only a small negative correlation between digital technology use and adolescent wellbeing.
Haidt and Twenge lauded the paper as “impressive and important,” but argued that it failed to disentangle general digital technology use from the specific problem they cite: social media use, in excess, primarily among teenage girls. Orben et al.’s most recent paper remedies this to some extent, focusing on social media in particular and separating boys and girls, but Haidt and Twenge object to the use of “linear r to examine data that is curvilinear.” That is, because the correlation between social media and wellbeing follows a J-curve (with no use and excessive use both being correlated to lower wellbeing), measuring the correlation linearly may wipe out the effect.
However the results turn out, the debate has become a powerful testament to the importance of criticism and open science. As Haidt and Twenge write, “whenever a new fad or technology sweeps through the child or teen population, stories get written about how the new trend is harming children.” Disentangling legitimate fears from moral panics depends upon high quality research, which in turn depends on methodical efforts to mitigate bias and seek out disagreement. For a topic as fraught as this, the very liveliness of the debate is a hopeful sign of progress.