Sierra Leone: AI Gemini Transforms Math Education

A Controlled Trial in Sierra Leone

A randomized controlled trial was conducted in Sierra Leone to assess the impact of AI on mathematics learning. This project, carried out in partnership with Fab AI and supported by the Ministry of Education, involved 1,763 secondary school students across 12 schools in the Port Loko district. For eight weeks, the guided learning tool Gemini was used to measure student progress. Conrad Sackey, Minister of Basic and Senior Secondary Education, expressed satisfaction with the results, emphasizing the importance of rigorous evaluation of educational innovations.

AI as a Teacher's Ally

The results of this trial show that AI can play a crucial role as a pedagogical partner. Rather than replacing teachers, the Gemini tool was designed to extend their reach and enhance teaching effectiveness. This study is part of a broader effort to establish a body of evidence on the impact of AI in education. The goal is to demonstrate how well-designed AI can support teachers and improve learning outcomes.

Preserving Critical Thinking

A common concern regarding generative AI is that it could become a shortcut for students, hindering their development of critical thinking. Guided learning with Gemini addresses this concern by focusing on understanding rather than providing direct answers. Data collected in Sierra Leone shows that this approach is effective. Out of more than 113,000 analyzed interactions, 91.4% enabled students to develop conceptual understanding rather than simply seeking solutions. Gemini posed supportive questions in 76% of its messages, providing direct solutions in only 2% of cases. This Socratic method ensures that the cognitive effort remains with the student.

The Central Role of Teachers

The success of this trial relies on close collaboration between AI and teachers. The latter designed the lessons, defined objectives, and facilitated classroom discussions. Teachers also reported that Gemini supported their own professional development. By using the tool to prepare their lessons, they discovered new ways to explain concepts like fractions. Many noted a transition from their role as "lecturers" to that of "facilitators," moving around the classroom to support students in their learning journey.

Significant Quantitative Results

The quantitative results of the trial are impressive. Students using guided learning recorded a gain of +0.258 standard deviations in their mathematics scores compared to the control group. This equates to approximately 1.2 to 1.7 years of typical learning progress in just eight weeks. Students whose teachers integrated Gemini into about half of their lessons experienced even greater gains, ranging from 1.8 to 2.5 years of progress. Student engagement was also remarkable: 69% met or exceeded usage goals, far surpassing the usual 5% rate for voluntary educational technologies.

Behavioral Change and Increased Engagement

Beyond the numbers, the trial revealed a profound change in student behavior. They expressed increased interest in mathematics and actively engaged in learning beyond regular instruction. Over time, their questions and conversations became more learning-oriented, with an increase in skill-development-focused questions from 68% to 90% between the first and last week of the trial. Questions seeking solutions decreased from 25% to 10%, indicating that students were seeking to understand how to arrive at answers rather than simply obtaining them.

Towards More Equitable Education

While the results are promising, they also highlight the challenge of the achievement gap. Students with already strong math skills benefited the most from the tool. This underscores the need to develop tools that maximize gains for students who need them the most. Expanding trials to other countries and exploring areas such as metacognition and relational intelligence are planned to ensure equitable learning opportunities for all. By combining the relational approach of a teacher-led classroom with the personalized capabilities of AI, it is possible to create a more inclusive and effective learning environment.

A Global and Future Perspective

To better understand the impact of guided learning, a series of other randomized controlled trials is underway globally. These efforts aim to build a transnational body of evidence that will support the responsible development of AI in education. A manual on the randomized controlled trial approach is also being published to assist others in conducting similar studies. Support for the Global Alliance for AI in Learning (GAILA) is another commitment to accelerate these initiatives.

Conclusion: A Persistent Challenge

Although the Gemini AI has shown promising results, the challenge of the achievement gap persists. Students with better initial skills have reaped the most benefits, highlighting the need for solutions that assist those who need it most. The future of these trials includes exploring metacognition and relational intelligence for a more holistic understanding of learning. By combining traditional teaching with the capabilities of AI, the goal is to ensure that technology serves as a bridge to meaningful learning opportunities for all.