DoodleLearning Adaptive Platform — A Randomised Evaluation Across 70 Belizean Primary Schools

In November 2024, Belize's Ministry of Education, Culture, Science and Technology partnered with the Inter-American Development Bank (IDB) to distribute Chromebooks preloaded with DoodleLearning's AI-adaptive platform to primary schools nationwide, aiming to reach more than 6,000 students. The platform's Proxima algorithm personalises maths and English exercises to each student's zone of proximal development.

Innovations for Poverty Action (IPA), working with IDB researcher Emma Naslund-Hadley and IPA's Juan Hernandez-Agramonte, set out to independently test whether the adaptive platform improved learning outcomes at national scale, running a randomised controlled trial across 70 schools and roughly 1,900 students aged 10-12 that compared DoodleLearning users to a randomly assigned comparison group.

The RCT found no statistically significant improvement in maths test scores, and only a modest improvement in English scores among students with complete usage data. Active users reported greater maths self-efficacy and more positive views of technology access, but girls reported less equitable access to devices at home, and teachers raised usability complaints.

IPA's own framing was blunt: technology alone is not enough. Without deeper curriculum integration, sustained teacher training, reliable infrastructure and an explicit equity lens, an off-the-shelf adaptive platform does not reliably move learning outcomes, even when rolled out at national scale.