CodeWalkthrough
Students explain their own code, by text or voice, until they show they understand what the assignment was meant to teach.
Working code is not proof of understanding. In large programming courses, students can submit a correct assignment without being able to explain how it works, and generative AI has widened that gap. An oral code walkthrough reveals what a student actually understands, but no instructor can hold one with every student in a class of hundreds.
CodeWalkthrough brings the walkthrough to scale. Instructors list the intended learning objectives (ILOs) for an assignment and provide its files and details. Students upload their code and explain their understanding of the relevant topics by text or voice, with follow-up questions and conversation until they have demonstrated the full set of objectives.
The design draws on learning-sciences research and on the lab’s VoiceEx work on spoken self-explanation: productive friction that asks students to think before they are helped, adaptive supports such as hints, and prompts to reflect and self-explain. Each walkthrough ends with post-assessment questions, giving instructors and researchers a measure of what students learned.
People
CORE Lab
- Angela Zavaleta BernuyCo-Director, CORE Lab; Assistant Professor, Computing and Software, McMaster University
- Michael LiutCo-Director, CORE Lab; Associate Professor, Teaching Stream, University of Toronto Mississauga
- Khayla SalimUndergraduate Student, Computer Science, McMaster University
- Deluckshan MurugesuM.Eng. Student, McMaster University
Papers
Papers will appear here as the project publishes.