A year-long panel of 1,966 adults finds shifts in social tech use do not forecast life satisfaction

The design turns on two questions that survey work usually collapses into one. Whether people who use a platform more than others are less satisfied with their lives is a comparison between people ^[10]. Whether one person grows less satisfied after stepping up their use is a comparison within that person ^[10]. "The second question is the one that gets us closer to cause and effect," said Kostadin Kushlev, co-first author of the paper ^[11][16].

Five rounds three months apart put twelve months between a respondent's first survey and their last ^[1]. That leaves four consecutive three-month transitions per person ^[3], and up to 9,830 person-wave observations across the 1,966 respondents ^[2].

The statistics are Bayesian. A conventional null reports only that the data missed a threshold; Kushlev said their approach can go further. It lets them say whether the effect credibly exists, he said ^[12]. The phys.org write-up does not report effect sizes or the widths of those credible intervals ^[22]. Those bounds are what would tell you how small an effect this panel can exclude.

The between-person associations did not vanish. Kushlev said they did find that people who use TikTok or YouTube more report lower life satisfaction, but that associations of that kind could be due to third factors like age ^[13]. Those are the correlations most cross-sectional studies report, by his account ^[23].

What the panel measured is how often adults said they used each of ten technologies, and their life satisfaction on an established three-item scale ^[5][6]. Frequency is a coarse input, and a global satisfaction judgement three months later is a coarse output ^[7]. An adolescent sample, a daily diary, or a clinical symptom measure would be different studies, and this one does not stand in for them.

The team also checked whether answers differed by age, gender, race, education and income, and said it chose a method that separates a real effect from the trivial noise that reaches significance in any large sample ^[17][18]. The subgroup breakdowns came back the same as the whole ^[9].

Kushlev put the state of the literature this way: "Different researchers have looked at the same data and come to opposite conclusions" ^[14]. On the framing of the question itself, he said: "Asking whether social media is good for well-being is a bit like asking whether eating is good for your heart. Different foods do very different things, and the effects also depend on the person and other factors" ^[15].