Andrew Przybylski
Andrew Przybylski is a psychologist at the Oxford Internet Institute, where he is director of research, and one of the small number of people whose work has genuinely changed what can responsibly be said about screens and young people. His method is consistent: very large samples, transparent analysis, and a refusal to report one specification when hundreds are defensible. The findings that come back are consistently small, and one of them is not merely small but shaped: wellbeing rises with moderate technology use before it falls, a pattern he and Weinstein named the Goldilocks hypothesis. He is first author on that paper, tested across 120,115 adolescents, and second author on the specification curve analysis that is usually credited to him alone.
Who they are
Przybylski's contribution is best understood as two findings and a method.
The first finding is the Goldilocks hypothesis, from Przybylski and Weinstein (2017) in Psychological Science, a study of 120,115 adolescents. The relationship between screen use and wellbeing is not a straight line downwards. Moderate use is associated with slightly higher wellbeing than no use at all, and the association only turns negative at high levels. The upside-down U is the part almost always lost when the research reaches a newspaper, because it does not support either of the two positions the public argument has room for. Just as usefully, the paper puts the size of the negative association at the far end into terms that mean something: at exceptional levels of use, the association is roughly a third as bad as missing breakfast or not getting eight hours of sleep. Those are not trivial things, and they are also not things anyone writes legislation about.
The second is Orben and Przybylski (2019) in Nature Human Behaviour, the specification curve analysis across 355,358 people that put the share of variation in adolescent wellbeing explained by digital technology use at no more than 0.4 per cent, slightly more than the association for regularly eating potatoes. Amy Orben is first author on that paper. It is routinely attributed to Przybylski, including by Nir Eyal and by this publication's own working notes until 1 September 2026. None of that is Przybylski's doing, and correcting it takes nothing away from him, since his own first-authored work carries the same argument on its own.
The method is the durable part. Both papers are demonstrations of how much of a published result is a product of the analyst's choices rather than the world's. Give a large survey dataset to enough researchers and it will produce alarming headlines and reassuring ones from the same numbers, without anyone doing anything improper: the choices about how to measure use, how to measure wellbeing, and what to control for are all arguable, and they all move the answer. Reporting the full distribution of defensible analyses instead of one of them is a replication-crisis reform wearing different clothes, and it is why these particular numbers have held up while more dramatic ones have not.
There is a small irony in the record. A 2013 paper by Przybylski and Weinstein on the presence of a phone during conversation appears in Nicholas Carr's own endnotes in The Shallows, which is to say the writer making the strongest case for harm and the researcher whose numbers most constrain that case are reading some of the same evidence and arriving in different places. On the replication record as it stands, Przybylski's findings have held better. The Goldilocks pattern and the breakfast comparison are still there. Two of the studies Carr leaned on hardest are not.
Key papers
Przybylski, A. K., & Weinstein, N. (2017), "A large-scale test of the Goldilocks hypothesis," Psychological Science, 28(2), 204-215. n = 120,115. Moderate digital engagement is associated with slightly higher wellbeing than none, with the association turning negative only at high levels of use.
Orben, A., & Przybylski, A. K. (2019), "The association between adolescent well-being and digital technology use," Nature Human Behaviour, 3, 173-182. Specification curve analysis, n = 355,358, at most 0.4 per cent of the variation in wellbeing explained. Second author; Amy Orben is first author.
Key ideas
The Goldilocks hypothesis: the relationship between screen use and adolescent wellbeing is an inverted U, not a slope, and moderate use sits above abstinence.
Effect sizes worth carrying: at exceptional levels of use, the negative association is about a third as large as missing breakfast or losing a full night's sleep.
Analytic flexibility is what lets the same dataset generate opposite headlines. Reporting every defensible specification is the fix.
Very large samples make statistical significance nearly automatic, which is why the argument has to be conducted in effect sizes.
Concepts on this site
Find them elsewhere
Bluesky and Scholar are linked from his OII profile and his own site przybylski.xyz. The X handle is identical to the confirmed Bluesky one and the account carries his name and OII work, so it is confirmed by content rather than by a direct link. LinkedIn states his Oxford chair
Checked 2026-09-10. Handles and affiliations go out of date, so treat this block as a snapshot rather than a standing fact.
Sources
- Przybylski, A. K., & Weinstein, N. (2017), "A large-scale test of the Goldilocks hypothesis," Psychological Science, 28(2), 204-215.
- Orben, A., & Przybylski, A. K. (2019), "The association between adolescent well-being and digital technology use," Nature Human Behaviour, 3, 173-182.
- Carr, N. (2010), The Shallows: What the Internet Is Doing to Our Brains, W. W. Norton. Tenth-anniversary edition with new afterword, 2020. (Cites Przybylski and Weinstein on phone presence and conversation.)
- Eyal, N. (2019), Indistractable: How to Control Your Attention and Choose Your Life, BenBella Books. (Attribution of the 2019 paper to its second author.)