The empirical association between adolescent social media use and mental health outcomes is small and methodologically fragile; the evidence does not warrant attributing the post-2012 youth mental health decline principally to smartphones, and the popular synthesis on this question has run ahead of its data.

Odgers and her co-authors have spent more than a decade working directly with the kind of datasets and analytic methods this debate hinges on, and they keep arriving at the same conclusion: the associations are real but small, the methods are more fragile than the popular synthesis lets on, and the cross-national timing pattern looks cleaner from a distance than it does once you break it apart by country and metric. The expert_opinion scheme licenses giving weight to that judgment in the specific sub-area where she has done primary methodological work. Her position isn't dismissive — she takes the rise in adolescent distress seriously and isn't denying it's happened — it's about what the existing data can and can't carry. Specification-curve analysis, which runs the same association across thousands of plausible model choices and looks at the distribution rather than picking one specification to publish, returns effects in the range of 0.4% of variance: not nothing, but not the kind of effect size that typically supports a "principal cause" claim. And several drivers that co-vary with smartphone adoption — declining sleep, less unsupervised face-to-face time, changes in clinical screening practice — can't be cleanly separated from social media in the available studies.

• Specification-curve analyses (Orben & Przybylski 2019 and successors) systematically vary analytic choices across thousands of plausible model specifications and find that the median association between adolescent digital technology use and well-being explains less than half a percent of variance — comparable in magnitude to associations with eating potatoes or wearing glasses.
  Evidence for this premise (5)

  Orben & Przybylski — Adolescent well-being and digital tech use (Nat Hum Behav 2019)

  https://www.nature.com/articles/s41562-018-0506-1

  Orben & Przybylski — Adolescent well-being and digital tech use (Nat Hum Behav 2019)

  https://www.nature.com/articles/s41562-018-0506-1

  Orben & Przybylski — Adolescent well-being and digital tech use (Nat Hum Behav 2019)

  https://www.nature.com/articles/s41562-018-0506-1

  Orben & Przybylski — Adolescent well-being and digital tech use (Nat Hum Behav 2019)

  https://www.nature.com/articles/s41562-018-0506-1

  Orben & Przybylski — Adolescent well-being and digital tech use (Nat Hum Behav 2019)

  https://www.nature.com/articles/s41562-018-0506-1
• Candice Odgers is a developmental psychologist at UC Irvine, past president of the Society for Research on Adolescence, and a methodological specialist in adolescent well-being research; her 2024 *Nature* review of the same evidence base concludes that the data do not support a primary causal role for smartphones.
• TODO: a single claim summarizing the set of plausible confounders (declining sleep, less unsupervised face-to-face time, changes in clinical screening practice) that the available studies cannot cleanly separate from social-media exposure.

Critical questions are the challenges this argument’s reasoning pattern must withstand.

• Is E an expert in D?
  Answer

  The CQ here isn't really "is Odgers credentialed?" — she obviously is. The harder version, given that Haidt is also credentialed and reaches the opposite conclusion, is: what licenses giving more weight to her judgment on this specific sub-question? The answer is methodological proximity. Odgers' primary research output for the past fifteen years has been exactly the kind of study this debate turns on: large-N longitudinal analyses of adolescent technology use and well-being, including direct authorship and commentary on the specification-curve methodology that has reshaped the field. Haidt's role in this debate is closer to synthesis and interpretation — he's reading the literature, weighing it, presenting a thesis to a general audience. That's a real and valuable activity, but it isn't the same as generating the underlying analyses and contesting their analytic choices. When the question is the narrow methodological one — "do these specific datasets and these specific methods support causal inference at the strength being claimed?" — the expert who works primarily inside those datasets and methods is the one whose judgment the expert_opinion warrant most directly licenses. This isn't a knockdown. Haidt has read the literature deeply, his counterarguments are reasoned rather than reflexive, and on broader questions about adolescent culture, attention, and norms his expertise is at least as relevant as Odgers'. The point is that the narrower the question gets — and the question this argument turns on is narrow, about what the existing evidence base warrants — the more her methodological proximity matters. On that question, the expert appeal does the work the scheme says it does.

These are challenges this argument’s reasoning pattern must still withstand. Answering them on Isonomia strengthens the argument.

• Do experts in D disagree on φ?Answer under review
  Draft answer

  TODO: address whether Odgers' position represents a methodological consensus among adolescent-well-being researchers, or whether it is itself contested within the field.

• Is E biased?Open
• Is E’s assertion based on evidence?Open

Cited by 3 · 1 contest