AI, implementation and academic outcomes in low-resource settings: Experimental evidence from Ghana

We propose a small scale implementation study of Rori, an AI-powered math tutor accessible via WhatsApp, and its impact on the math performance of students in low-resource settings. Previous research conducted at Rising Academies schools found a moderate to large effect size, but the ambition is for Rori to scale to more students. Thus, this study will focus on the effects of implementation fidelity on learning gains in other low-cost private schools and government run schools in Ghana, compared to Rising Academies schools. We plan to recruit 24 schools that will be split into two conditions with students in the treatment group given access to Rori for one hour a week to independently study math. Students will be assessed using items from TIMSS and their learning gains calculated and measures of fidelity will be collected. We hypothesize that Rori will be effective in improving math learning gains among students in the treatment group compared to the control, but the size of learning gains will vary according to implementation fidelity. Learnings from this pilot will inform a future larger randomized evaluation, as well as have broader relevance and applications for policy and decision-making beyond this test case.