Edo State AI Tutoring After-School Programme — World Bank RCT Finds Two Years' Learning in Six Weeks

In June–July 2024, World Bank researchers partnered with the public secondary-school system in Edo State, southern Nigeria, to test whether generative AI could accelerate learning in an under-resourced setting.

Test whether structured, teacher-supervised use of a generative-AI chatbot could accelerate learning gains for first-year senior-secondary students in English, in an after-school setting.

About 800 students across roughly nine public schools attended six weeks of after-school English sessions, twice a week. Each session opened with a teacher-led introduction tied to the curriculum, followed by structured, reasoning-oriented interaction with ChatGPT delivered via Microsoft Copilot; teachers stayed in the room throughout to guide use and catch AI errors. Unreliable power and internet access required upfront basic digital-literacy training before the AI sessions could begin.

A randomized evaluation found an overall learning gain of 0.3 standard deviations (0.24 SD in English specifically) — equivalent to roughly 1.5 to 2 years of typical schooling progress, and ahead of about 80% of documented education interventions worldwide. Average exposure was 13 hours over six weeks, and gains rose with attendance. Female students and students who started furthest behind improved the most, narrowing rather than widening the pre-existing achievement gap, and students also scored better on standardised exams beyond the material directly covered.

The World Bank team is explicit about the limits: the trial ran after school rather than during the regular school day, lasted only six weeks, and tested English only, so durability, subject transfer and full classroom integration remain open questions.