Kwame for Science — AI Teaching Assistant for West African Science Students

Kwame for Science is an AI teaching assistant built by the Ghanaian ed-tech venture SuaCode.ai (creator of the original Kwame AI coding-education chatbot) to answer West African secondary students' questions on the Integrated Science syllabus tested in the West African Senior Secondary Certificate Examination (WASSCE). Students submit questions in natural language through a web app and receive curated passages and related past-exam questions as answers, rather than open-ended generative text.
An 8-month real-world deployment reported in a peer-reviewed AIED 2024 evaluation logged 750 users across 32 countries (15 of them in Africa) and roughly 1,500 questions asked. On a held-out set of 109 questions, Kwame's top-3 suggested answers included a useful response 87.2% of the time, though top-1 accuracy (the single best answer) was materially lower — a limitation the authors report openly rather than obscuring.
A follow-up system, Kwame 2.0, extended the approach to coding education and was studied over 15 months across 35 African countries with 3,717 enrollments, using human facilitators and peers to catch and correct the AI's errors — evidence that the team treats accuracy limits as an ongoing design constraint, not a solved problem.

Read the full analysis: https://suacode.ai/suacode-ai-launches-kwame-ai-in-beta-an-ai-powered-web-app-that-provides-instant-answers-to-students-science-questions/