Measuring the impact of learning with AI in Sierra Leone and beyond

2026-06-09 · Source: Google DeepMind News · Field: Education & Learning — Educational Technology (EdTech), Academic Research & Higher Education, K-12 Education & Child Development · Depth: Intermediate, short

Summary

A randomized controlled trial (RCT), conducted over eight weeks in partnership with Fab AI and the Sierra Leone Ministry of Education, evaluated the impact of Guided Learning in Gemini on math progress. The study involved 1,763 junior secondary students across 12 schools in Sierra Leone's Port Loko District. Results showed students using Guided Learning achieved a gain of +0.258 standard deviations in math scores, representing 1.2 to 1.7 years of typical learning progress. Students whose teachers integrated Gemini for approximately 12 hours saw even higher gains, roughly 1.8 to 2.5 years of progress. Analysis of over 113,000 interactions revealed 91.4% focused on conceptual understanding, with Gemini providing scaffolding questions in 76% of responses. Teachers remained central, designing lessons and using the tool for professional development. The trial also highlighted an achievement gap, with stronger students benefiting most.

Key takeaway

For Directors of AI/ML in Education considering large-scale deployments, this trial demonstrates that AI can substantially improve learning outcomes and student engagement. You should prioritize AI tools designed for "Socratic" scaffolding that build conceptual understanding, not just provide answers. Ensure your implementation strategy keeps teachers central, empowering them to design lessons and integrate AI effectively, while also addressing potential achievement gaps.

Key insights

AI, when designed for "Socratic" scaffolding, significantly improves student math learning and engagement without replacing teachers.

Principles

• AI can augment teachers, not replace them.
• Prioritize understanding over direct answers.
• Teacher-led integration maximizes AI impact.

Method

The Guided Learning approach uses AI to pose scaffolding questions, encouraging students to build conceptual understanding rather than seeking direct solutions. Teachers design lessons and facilitate discussions, integrating the AI tool for student support.

In practice

• Implement "Socratic" AI for conceptual learning.
• Provide teacher training for AI integration.
• Conduct RCTs to validate educational AI tools.

Topics

• Guided Learning
• Educational AI
• Randomized Controlled Trials
• Math Education
• Teacher Professional Development
• Learning Outcomes

Best for: AI Scientist, Research Scientist, Policy Maker, Director of AI/ML

Related on AIssential

• We designed an AI tutor that helps college students reason rather than give them answers — AI designed as a Socratic tutor, combined with peer discussion, enhances student learning outcomes.
• The next phase of OpenAI’s Education for Countries — Responsible, government-led research partnerships are crucial for safely integrating agentic AI into education to empower students.
• More and more teachers and students are using AI – even though it might do more harm than good — Rapid AI adoption in K-12 education outpaces policy and evidence, raising concerns about safety and learning outcomes.
• Our AI Literacy Day recap: putting educators in the lead — Empowering educators with AI tools fosters innovative teaching and significantly streamlines administrative tasks.
• 🎧 Inside an AI High School, Through the Eyes of a 17-Year-Old Founder — AI can transform education and youth perspectives on learning, social interaction, and career paths.
• Can AI Tools Transform Low-Demand Math Tasks? An Evaluation of Task Modification Capabilities — AI tools show moderate success in upgrading low-demand math tasks, with generative modification distinct from classification.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Google DeepMind News.