Scaling the Oral Exam — UC San Diego's NSF-Funded Project Runs 8,500 Video Oral Exams Across 10 Engineering Courses

As generative AI made take-home and online assessments unreliable, Huihui Qi, an associate teaching professor in UC San Diego's Department of Mechanical and Aerospace Engineering, led an NSF-funded three-year project -- 'Improving the Conceptual Mastery of Engineering Students in High Enrollment Engineering Courses through Oral Exams' -- to test whether short oral exams could be administered at the scale of large lecture courses rather than only in small seminars.

The project ran across 10 courses with seven participating professors, administering roughly 8,500 short oral exams (about 15 minutes each, mostly conducted over video). Because oral exams are labour-intensive, the team built a two-part teaching-assistant training programme -- video-based asynchronous behavioural training plus course-specific, instructor-led technical training -- so TAs could conduct exams consistently and fairly at scale rather than relying solely on faculty.

UC San Diego Today reported the project's own framing of the results as improving both engineering student performance and motivation, and the work has drawn coverage from the Associated Press, the Wall Street Journal and EdSurge, plus invitations for Qi to run faculty workshops on scaling oral exams at other institutions -- a signal of transferability beyond the original courses.

The publicly available reporting does not include a published effect size, a control-group comparison, or independent replication outside UC San Diego, so the evidence here should be read as a well-documented, at-scale single-institution project rather than a rigorously controlled trial.