NigeriaEducation2024Edo State public schools and the World Bank
After-school generative AI tutoring pilot
Six weeks of teacher-guided AI tutoring raised Nigerian students' English scores by about 0.3 standard deviations
Problem
Secondary schools in Edo State, Nigeria, face large classes and low learning levels, and many students had never used a computer. The state and the World Bank wanted to test whether a general-purpose AI chatbot could act as an affordable tutor.
Approach
In June and July 2024, randomly selected first-year secondary students in Benin City attended twice-weekly after-school sessions over six weeks in school computer labs. They used a GPT-4-based chatbot (reported as Microsoft Copilot) as a virtual English tutor, with teachers guiding prompts and activities aligned with the curriculum. World Bank researchers compared them with students who were not selected, using pen-and-paper tests.
Results
Takeaway
Teachers led the short program with a general-purpose tool, without a purpose-built education product; attendance tracked results, and girls gained more than boys. The pilot was small and short, and its authors said scaling needed further research.
Categories
- Size
- Large enterprise
- Industry
- Education
- Business function
- Production and quality
- Innovation type
- Product and service innovation
- AI technology
- Generative AIConversational and voice AI
- Adoption
- Government support program
- Evidence
- Independent study
Sources
Source summaries
The World Bank research team describes the six-week after-school pilot run in June and July 2024 in Edo State. It reports an effect of about 0.3 standard deviations, which it equates to roughly two years of typical learning, and says the result beat 80% of interventions in a database of randomised trials in developing countries. Gains reached end-of-year exams, were larger for girls and grew with attendance.
The Decoder reports that students used Microsoft Copilot in twice-weekly, teacher-guided sessions and scored 0.3 standard deviations above the control group. It quotes Ethan Mollick's caution that the pilot was limited in time and scope and that teachers likely played a central role. It notes that the long-term effects are unknown.
Full details are in the original sources.
