Kwame: A Bilingual AI Teaching Assistant for Online SuaCode Courses
Project Overview
The document highlights the innovative use of generative AI in education through the introduction of Kwame, a bilingual AI Teaching Assistant specifically designed for online coding courses like SuaCode. Utilizing a Sentence-BERT based question-answering system, Kwame effectively answers student inquiries in both English and French, thereby improving the learning experience in large online classes, particularly during the challenges posed by the COVID-19 pandemic. By alleviating the workload on human facilitators, Kwame enhances engagement and support for students, making coding education more accessible. The initiative aims to democratize coding education across Africa, leveraging the widespread availability of smartphones to reach a broader audience and foster skill development in technology. Overall, the implementation of Kwame represents a significant advancement in using AI to support and enrich online education, addressing both logistical challenges and educational equity.
Key Applications
Kwame - Bilingual AI Teaching Assistant
Context: Online coding courses for students from 42 African countries, primarily Anglophone and Francophone.
Implementation: Developed using Sentence-BERT for real-time question answering, trained on course materials and past student questions.
Outcomes: Provides quick and accurate answers, potentially improving learning experiences and reducing the burden on human TAs.
Challenges: Scalability of support, accuracy of responses to diverse student inquiries, and handling of noisy real-world questions.
Implementation Barriers
Technical Barrier
Difficulty in providing accurate answers to varied and noisy student questions.
Proposed Solutions: Future work will aim to clean questions and improve semantic search approaches.
Scalability Barrier
Human support is limited and not scalable as student numbers increase.
Proposed Solutions: Implementing AI TAs like Kwame to provide automated support.
Project Team
George Boateng
Researcher
Contact Information
For information about the paper, please contact the authors.
Authors: George Boateng
Source Publication: View Original PaperLink opens in a new window
Project Contact: Dr. Jianhua Yang
LLM Model Version: gpt-4o-mini-2024-07-18
Analysis Provider: Openai