Amy Orben
Amy Orben is an experimental psychologist at the MRC Cognition and Brain Sciences Unit in Cambridge, and the first author of the single most useful study in the screen-time argument. Orben and Andrew Przybylski (2019), published in Nature Human Behaviour, ran a specification curve analysis across three large datasets covering 355,358 people and found that digital technology use explains at most 0.4 per cent of the variation in adolescent wellbeing. To give that number a scale a reader can hold, they observed that it is only slightly more than the association for regularly eating potatoes. The method matters more than the headline, and the headline is very widely quoted, usually with her name left off it.
Who they are
Orben works on adolescent wellbeing and digital technology, and on the methodological question that sits underneath it: how much of what a study finds was decided by the researcher rather than by the data. The 2019 paper is the clearest answer anyone has given to that question in this field, which is why it has outlasted almost everything published around it.
The problem it was built to solve is analytic flexibility. A large survey dataset on teenagers, screens and wellbeing contains hundreds of defensible ways to ask the same question. You can define technology use as hours, as any use, as use after a certain hour of the evening. You can define wellbeing as depressive symptoms, as life satisfaction, as self-esteem, as a parent's report rather than the child's. You can control for family income, or not. Each of those choices is arguable, and researchers have to make all of them before they see a result. The consequence is that a single dataset can yield an alarming headline or a reassuring one, and the ordinary way of reporting research, one analysis presented as the analysis, shows the reader nothing about which of those two happened.
Specification curve analysis, which Orben applied here rather than invented, deals with this by refusing to choose. Instead of running one analysis and reporting it, you enumerate every defensible way of specifying the question, run all of them, and plot the whole distribution of results. In the 2019 paper that meant tens of thousands of specifications. What comes back is not a number but a shape: how large the effect is across the full range of reasonable choices, how much of the spread is negative, how much is positive, and how far out on the tail you would have to go to find the striking result someone else has reported. It makes cherry-picking visible, and it makes the 0.4 per cent figure far more trustworthy than any single striking number could be, because it is not one finding but the centre of every finding the data can honestly support.
The result itself is deliberately unexciting. Technology use is associated with adolescent wellbeing negatively, and the association is real, and it is tiny. The potato comparison, and the others in the same paper, exist to stop readers from mistaking statistical significance in a sample of 355,358 for practical importance. With samples that large, almost everything is significant, so effect size is the only question worth asking.
There is an attribution problem attached to this work, and this site has been part of it. The 2019 paper is routinely credited to Przybylski, who is the second author. Nir Eyal credits it to him. This publication's own working notes credited it to him until 1 September 2026, when the error was caught during the reporting of "Both Sides of the Screen-Time Argument". It is the fourth time we have found the same mistake in our own files, always the same shape: the more famous of two names absorbs the credit for the work of the first author. It is worth saying plainly rather than quietly fixing, because the correction only sticks if the pattern is named.
Key papers
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 across three datasets, n = 355,358. Digital technology use explains at most 0.4 per cent of the variation in adolescent wellbeing, slightly more than regularly eating potatoes.
Key ideas
Specification curve analysis: run every defensible version of an analysis and report the whole distribution, rather than running one and reporting it as though it were the only one available.
Analytic flexibility is the mechanism behind contradictory headlines from the same data, and it does not require anyone to behave dishonestly. Ordinary defensible choices are enough.
In samples of hundreds of thousands, statistical significance is close to guaranteed, so the only useful question is how large the effect is.
On the current best evidence, the effect of digital technology use on adolescent wellbeing is negative, real and negligible in size.
Concepts on this site
Find them elsewhere
LinkedIn and Scholar are linked from her own site amyorben.com, the X handle from her lab site orben.group, and the Bluesky account states her MRC CBU group leader role
Checked 2026-09-10. Handles and affiliations go out of date, so treat this block as a snapshot rather than a standing fact.
Sources
- Orben, A., & Przybylski, A. K. (2019), "The association between adolescent well-being and digital technology use," Nature Human Behaviour, 3, 173-182.
- Eyal, N. (2019), Indistractable: How to Control Your Attention and Choose Your Life, BenBella Books. (Attribution of the 2019 paper to its second author.)