Early Exposure to AI Training and Students’ Educational Trajectories in Ethiopia

Policy Context

Education systems in East Africa face challenges including low student engagement, high dropout rates, poor learning outcomes, and significant teacher attrition, particularly in rural and underserved areas. These issues occur alongside rapid population growth and limited access to digital tools. While some emerging economies have begun integrating artificial intelligence into their education systems, East African countries have had limited adoption of these technologies due to resource constraints and infrastructure limitations.

There is growing recognition in Ethiopia of the need to modernize the education sector and improve educational quality. This study examines whether AI-powered learning tools and teacher training programs can address some of the persistent challenges in the education system. The study focuses on rural areas in Benishangul-Gumuz, where the Pharo Foundation runs multiple schools.

The research is relevant to current policy discussions in Ethiopia regarding digital literacy, teacher professional development, and curriculum reform.  Findings may ultimately inform decisions about whether and how to integrate AI tools into education programs.

Study Design

This pilot study employs a randomized controlled trial (RCT) to examine the effects of early exposure to AI-based learning tools and teacher capacity-building on students’ educational trajectories. The study focuses on upper-primary and lower-secondary school students in two Pharo Foundation schools and one public school in Benishangul-Gumuz. These schools are located in rural areas representing underserved regions where educational resources and infrastructure are limited.

The intervention comprises two mutually reinforcing components. First, students in the treatment arm receive access to adaptive learning systems (M-Shule and/or iCog) designed for low-bandwidth contexts. These platforms deliver personalized lessons, assessments, and feedback aligned with the Ethiopian curriculum, enabling students to learn at their own pace in mathematics and English. Students also receive weekly AI literacy modules introducing basic AI concepts and practical applications. Second, teachers undergo structured training on integrating AI tools into instruction, including workshops on classroom data use, differentiated instruction, and digital pedagogy, supported by monthly follow-up sessions.

The study includes approximately 300 students equally allocated into treatment and control groups. Randomization is stratified by grade level to ensure balanced baseline characteristics. The control group continues with standard curriculum and traditional teaching methods.

The empirical analysis employs a Difference-in-Differences framework, comparing changes in student learning outcomes, engagement measures, and educational aspirations between baseline and endline across treatment and control groups. The platforms’ integrated data-collection features will complement primary survey and test-score data, enabling continuous evaluation of student learning trajectories throughout the intervention.

Results and Policy Lessons

Results are forthcoming.