While many fret over the proliferation of artificial intelligence (AI), the jury is still out regarding its benefits and drawbacks for K–12 education, and researchers are trying to catch up to the rapidly-expanding technology. The latest of two recent articles on the topic from the World Bank examines one education-related AI project in Africa.
In June and July 2024, 800 first-year students in senior secondary school in Nigeria (ages 16-17) were randomly assigned to attend after-school English classes in computer labs twice a week. Each session began with the teacher introducing the new weekly topic, followed by students interacting with the well-known Copilot generative AI tool from Microsoft, powered by ChatGPT. Teachers guided and mentored students throughout each session, which dealt with both grammar and writing tasks in English. Teachers also led brief reflection exercises at the end of each session. Importantly, a control group of students from the same schools who were not selected for the AI-powered sessions were also followed, although the articles discussing the intervention are light on details. (A proper report is promised soon, but until then, a video gives further detail on what the project looked like in practice.) After the six-week session, all students took a pen-and-paper test to assess their performance in three areas: English language skills, AI knowledge, and digital skills.
Students who participated in the AI-assisted classes performed better on their final tests to the tune of a statistically-significant 0.3 standard deviations. This is equivalent to nearly two years of typical learning in just six weeks. Compared to a database of other RCT education interventions conducted in the developing world, analysts say the AI intervention outperformed 80 percent of them. These include some widespread strategies like structured pedagogy and teaching at the right level. However, we don’t yet have a breakdown of how students did on the three individual parts of the test so it will be important to see the difference between English skills and the technological topics. Benefits accrued to all participating students, not just the highest achievers. And female students—whose success is an ongoing concern in African education—gained more compared to males. Attendance mattered greatly to the impacts, with every additional day of attendance resulting in statistically-significant improvements in learning outcomes. This is important, as attendance was inconsistent for many students due to factors beyond their control (family, weather, etc.). Making such classes fully virtual would reduce this concern, but home/community internet access would likely need to be addressed.
An earlier article published in September briefly describes seven takeaways from the same AI pilot project. On the upside, both students and teachers were highly engaged and reported wishing they had had even more time with the technology. (The introductory sessions were necessarily spent on housekeeping and training on the basics.) Teachers also reported a greater ability to immerse students in the English language using AI rather than other online (or written) resources. On the downside, teachers reported instances of hallucination (AI tools generating false responses and presenting them as facts) and student misuse of the technology, as well as the aforementioned attendance problems.
There’s a whole lot more we need to know about this experiment and others like it before any conclusions can be made beyond “this seems promising.” What kind of instruction did control group students get? How much human intervention is required to minimize AI problems and maximize student benefits? What kind of costs are required to reach the highest number of students? Are there other options besides Copilot that could work better? Until then, we remain at the “experts think” phase of technological innovation, sprinkled with a soupçon of “cautious optimism.”
SOURCE: Martín E. De Simone et al., “From chalkboards to chatbots in Nigeria: 7 lessons to pioneer generative AI for education,” World Bank blogs (September 2024).
SOURCE: Martín E. De Simone et al., “From chalkboards to chatbots in Nigeria: Transforming learning in Nigeria, one prompt at a time,” World Bank blogs (January 2025).
