AI-Powered Learning Boosts Math Outcomes for Students in Sierra Leone Trial

A recent randomized controlled trial (RCT) conducted in Sierra Leone has demonstrated the significant positive impact of AI-powered "Guided Learning in Gemini" on junior secondary students' math performance. The eight-week study, a collaboration between Google's Fab AI, the Sierra Leone Ministry of Education, Google.org, and the Gates Foundation, involved 1,763 students across 12 schools in the Port Loko District. The findings provide strong evidence that carefully designed AI can effectively augment traditional teaching methods and improve learning outcomes.

The "Guided Learning" tool is specifically engineered to foster deep conceptual understanding rather than merely providing direct answers. Analysis of over 113,000 student interactions revealed that 91.4% of conversations focused on building understanding, with Gemini responding through Socratic-style scaffolding questions in 76% of its messages. This approach ensures students engage in the necessary cognitive effort for deeper learning, counteracting concerns that generative AI might encourage shortcuts.

Teachers played a central role in the trial, designing lessons, setting objectives, and facilitating classroom discussions. They reported that Gemini not only supported student learning but also enhanced their own professional development, helping them discover new teaching methods and transition from "lecturers" to "facilitators." Student engagement was exceptionally high, with 69% meeting or exceeding usage targets, far surpassing typical voluntary educational technology adoption rates.

Quantitatively, students using Guided Learning achieved a gain of +0.258 standard deviations in math scores, equivalent to 1.2 to 1.7 years of typical learning progress within the short trial period. Classrooms where teachers integrated Gemini into about half their lessons saw even greater gains, up to 2.5 years of progress. Beyond scores, students showed a profound behavioral shift, reporting increased enjoyment in math and a sustained focus on skill-building queries over direct solution-seeking.

While the results are promising for leveraging AI to bridge educational gaps, the study also highlighted that students with stronger initial math skills benefited most, underscoring the need for tools that deliver the greatest gains to those who need them most. Google plans to expand these trials globally and continue publishing results to build a comprehensive, cross-country evidence base, aiming to inform responsible AI development across the learning ecosystem, particularly in contexts like Africa.