A randomised Peru trial found laptops raised use, not scores

The classroom note

One Laptop per Child promised that hardware alone could close an education gap: give a child a low-cost computer and let use follow. The programme, its own current site states, 'began at the Massachusetts Institute of Technology (MIT) in 2005' and has since 'provided more than 3 million educational laptops to children around the world,' including a large national rollout in Peru. The test of the theory came later, when researchers were able to compare schools that received the laptops with similar schools that did not.

What the evidence says

That comparison, published as a peer-reviewed study in the American Economic Journal: Applied Economics in July 2017, was 'a large-scale randomized evaluation of the One Laptop per Child program' across '318 primary schools in rural Peru,' measured after 15 months. The randomised design let the authors isolate the effect of the laptops themselves from other differences between schools that adopt new technology and those that do not. The intervention worked as hardware: it produced 'substantial increases in use of computers both at school and at home.' On the outcomes a school system cares about, the result was blunt: 'no evidence is found of effects on test scores in math and language.' The authors note 'some evidence, though inconclusive, about positive effects on general cognitive skills' — a hedged, secondary finding, not the study's headline result.

The implementation question

The gap the study points to is not access but instruction: a laptop that increases use without a curriculum, trained teachers or a specific instructional purpose attached to it increases screen time, which this trial's own measurement shows, without necessarily increasing the academic skills a maths or language test is built to detect. That is a mechanism question, not a hardware question: what was the laptop used for, during the school day, that a textbook or exercise book was not already doing.

What holds and what fails

What holds is the causal claim itself, because the randomised design supports it directly: in this evaluation, at this scale, over 15 months, laptops raised computer use and did not raise measured test scores. What fails is the premise the programme was built on, that access to a device would translate into learning gains on its own; a single large trial does not prove that no laptop programme anywhere can raise test scores, but it does show that this one, evaluated rigorously, did not, and any claim resting on device counts alone should be read against that result.

• Is a device programme's evidence about usage, or about a measured learning outcome?
• What instructional purpose, curriculum or teacher training was the device paired with, if any?
• Is inconclusive evidence for a secondary outcome being presented as the study's main finding?

The OLPC-Peru evaluation is the clearest large-scale precedent for a claim now made about AI access programmes: that putting a capable device or tool in a child's hands is, on this evidence, a plausible condition for learning gains but not by itself a sufficient one.