The social media panic is based on bad science

In his 1996 book, "The Vision of the Anointed", economist Thomas Sowell sketched out a pattern that many of the "crusading movements" of the 20th century have followed. First, they identify a "great danger" to society, followed by an "urgent need" for government action "to avert impending catastrophe."

A new book by psychologist and author Jonathan Haidt, The Anxious Generation, argues that the government must regulate social media because it's causing a teen mental health crisis. Haidt is, in many ways, a model researcher because of his rigor, transparency, and openness to dissent. On this issue, however, he fits neatly into Sowell's framework.

----

Those best equipped to get attention from the government and the media are the most "articulate" people, Sowell observes, and they often reference opaque studies without explaining them. And Haidt is certainly articulate—his book is well-written and filled with compelling insights. But he claims far too much certainty for his views, based on research that is mostly junk. And he advocates for restrictive government policies without doing the simple tests that might support or disprove their value.

Academic studies often make use of statistical techniques that are hard for the average person to decipher, which is a shame because "most published research findings are false," as Stanford's John Ioannidis argued in a 2005 paper. Ioannidis wasn't just referencing the many scandals of fabricated data, conscious or unconscious bias, and misrepresented findings. Even top researchers at elite institutions have been guilty of statistical malpractice. Peer review is worse than useless, better at enforcing conventional wisdom and discouraging skepticism than weeding out substandard or fraudulent work. Academics face strong pressure to publish flawed research. Few have the skill and drive to produce high-quality publications at the rate required by university hiring and tenure review committees. Even the best researchers resort to doing some easy, low-quality studies. Bad studies tend to be the most newsworthy and the most policy-relevant.

Many of the papers Haidt compiled contained coding errors, inappropriate statistics, and other issues. Most downloaded some data of little relevance—either cheap to generate, like surveying your sophomore psychology students, or data collected for a different purpose—and analyzed it with an off-the-shelf statistical approach.

Haidt cites 476 studies in his book that seem to represent an overwhelming case. But two-thirds of them were published before 2010, or before the period that Haidt focuses on in the book. Only 22 of them have data on either heavy social media use or serious mental issues among adolescents, and none have data on both.

Photos: Daniel Hambury/Eyevine, Mikkel Aaland/Free To Choose, E. Jason Wambsgans/TNS/Newscom, CNP/AdMedia/SIPA/Newscom, BONNIE CASH/UPI/Newscom, Richard B. Levine/Newscom, Howard Lipin/TNS/Newscom

Music: "Stream" by ANBR, "Empty Rooms" by Gal Lev, "Eclipse" by Yuriy Leontiev, "Curiosity" by Kevin Graham, "Discovery" by We Dream of Eden, and "Continent' by ANBR. Roadway by Ardie Son