Exploring a hybrid model where AI augments human Teaching Assistants to provide scalable, scaffolded programming feedback.
Lead PI:
Team Members:
Research Period:
Prof. Ben Leong Wing Lup
Umair Z. Ahmed, Shubham Sahai
2024 – 2025
About This Project
We investigated the use of AI to assist human tutors in providing feedback for coding assignments. In a randomised controlled trial with 185 CS1 undergraduates at IIT Kanpur, a large public University in India, we explored a hybrid model where Codaveri augments human Teaching Assistants (TAs) in providing programming feedback to students during a live lab session. Codaveri helped address educational scaling by acting as a scaffold, where AI assists with error identification while educators retain oversight to refine responses, flag inaccuracies, and contribute to the overall improvement of our system.
Codaveri hybrid feedback model augmenting human TAs

Ratings given by students for various feedback types (direct AI, augmented AI, or manual)

Publications: Feasibility Study of Augmenting Teaching Assistants with AI for CS1 Programming Feedback

Leave a Reply