Automated Oral Assessments at Scale for Safeguarding Academic Integrity

Generative AI has made it increasingly difficult for conventional programming assignments to establish whether submitted code reflects a student’s own understanding. The proposed project addresses this challenge through assessment redesign rather than AI detection. It will extend an existing prototype that automatically conducts structured oral assessments based on each student’s submitted program, asking them to explain, justify, and reason about their own implementation decisions.

The current system ingests programming submissions, generates personalised questions tied to each student’s code, collects spoken or written responses, and provides provisional evaluation against a transparent two-dimensional rubric: correctness (factual and technical accuracy of the response) and understanding (depth of reasoning, justification of design choices, and ability to explain consequences or alternatives).