Kwame for Science: AI Teaching Assistant Launched in West Africa

Africa faces a significant challenge with high student-to-teacher ratios, limiting students’ access to personalized learning support and making it difficult for them to get answers to their questions. To address this, we enhanced Kwame, our AI teaching assistant, adapting it specifically for science education, and deployed it as a web app.

Kwame for Science is designed to assist students by answering questions based on the Integrated Science curriculum of the West African Senior Secondary Certificate Examination (WASSCE). Powered by Sentence-BERT, the web app provides three detailed paragraphs as answers to science-related queries, along with a confidence score. Additionally, it displays the top five related past exam questions and their answers to reinforce learning.

During a 2.5-week real-world deployment, Kwame for Science demonstrated a top-3 accuracy of 87.5% (n=56), serving 190 users across 11 countries. This innovative solution has the potential to deliver scalable, cost-effective, and high-quality remote education, empowering millions of learners across Africa with accessible science education.

Source: arxiv