Google DeepMind: Gemini AI tutor gave Sierra Leone students 1.5 years of progress in 8 weeks

Google DeepMind: AI Tutor Gemini Gave Sierra Leone Schoolchildren 1.5 Years of Progress in 8 Weeks

Google DeepMind, together with Fab AI and Sierra Leone's Ministry of Education, completed a large-scale randomized controlled trial (RCT): 1763 middle school students studied mathematics for eight weeks using Guided Learning in Gemini. This is part of DeepMind's global program to study the real impact of AI on learning.

How the Experiment Was Designed

The study covered 12 schools in the Port Loco district. The participation of the state ministry of education ensured access to real classrooms and official policy-level support. The study was pre-registered — making it one of the most rigorous in EdTech.

Students did not simply ask Gemini for answers: the Guided Learning feature is specifically designed to encourage independent thinking. This approach is built on years of research from the LearnLM program. Analysis of 113,000+ dialogues showed:

• In 91.4% of conversations, students used Gemini to understand concepts rather than obtain ready-made answers
• In 76% of its messages, Gemini responded with clarifying questions — a "Socratic" approach
• AI provided ready-made solutions in only 2% of cases
• Key design principle: cognitive load remains on the student, not removed by AI

Results in Numbers

Students using Guided Learning showed a gain of +0.258 standard deviations in mathematics compared to the control group. In practical terms, this equals 1.2–1.7 years of learning progress in eight weeks. In classrooms where teachers incorporated Gemini into approximately half of lessons (target indicator — 12 hours over the period), the result was even higher: 1.8–2.5 years of learning progress.

Meanwhile, 69% of students met or exceeded target usage indicators — a striking contrast to the "five percent problem": typically, that many users remain with edtech tools voluntarily. In this study, the barrier was overcome more than tenfold.

Over time, students' conversations with Gemini became increasingly focused on learning: they less frequently sought hints and more often asked conceptual questions.

"We need to rigorously study the results of our innovations.

I am pleased that now we have solid evidence: carefully designed AI can improve learning outcomes in support of our teachers," — Conrad Seki, Minister of Basic and Secondary Education, Sierra Leone.

The Role of Teachers

The experiment was designed as a partnership, not a replacement. Teachers designed lessons, set goals, and led classroom discussions. Gemini took on the role of a personal tutor for each pair of students while the teacher moved about the classroom and supported those who were stuck.

In focus groups, educators noted that working with the tool affected their own preparation: using Gemini to develop lessons, they discovered new ways to explain long-familiar topics — for example, fractions. Many described a transformation in their role: from "lecturers," they became "facilitators."

DeepMind and Fab AI are releasing a methodological guide with protocols for other schools.

What This Means

This is one of the first large-scale randomized studies of generative AI in education in the Global South. It shows: properly designed AI tutoring does not replace teachers and does not open loopholes for cheating — it strengthens what educators already do and helps students think independently. For the EdTech industry, this is an argument in favor of "Socratic" design: ask questions rather than provide answers.