Unrestricted AI Access for Engineering Projects Coupled with Oral Interviews

This article is part of a series of case studies from McMaster instructors that explore perspectives and approaches for integrating or limiting the use of AI in learning environments.
–

Dr. Mike Welland from the Engineering Physics Program in the Faculty of Engineering utilized a combination of unrestricted AI use and oral interviews to assess students work in his Numerical Methods for Engineering course (3NM4/2NM4) course. Historically, part of the course has been assessed through completion of an assignment without the need for oral interviews. The capability for Generative AI to deliver precise, accurate answers that are not reliably detectable motivated a new method of assessing students actual understanding of the assignment material.

The Change

Dr. Welland did not feel that restricting students from using AI would be possible from an administrative perspective and that students will be required to use this technology in their future engineering employment; therefore, they should be encouraged to learn how to use it during their formal education. Furthermore, the course is conducted on the Google Colab Jupyter environment which removes accessibility barriers but directly integrates Google Gemini.

This motivated a shift in learning outcomes to emphasize conceptual fluency over rote learning. In practice, assignments were divided into written and oral components. While the written component is open to collaboration / AI, the oral component is 1-on-1 with a TA and secure. This new process was used for a class of approximately 100 students with 4 full-time Teaching Assistants.

The New Process

Assessment Process

• Students complete a written assignment in which they are able to collaborate or use AI freely.
• During in-person tutorials, the TAs conduct 1:1 interviews with each student in the class, approximately 5 min each. A invigilated written alternative is also provided for UDL purposes.
• During the interview, students must be able to recall and explain the content in their assignment. They must be able to justify the engineering decisions they made and explain how core engineering concepts work.
• Students are provided with a percentage for the written component (Score A) and another percentage for the oral component (Score B).
• The final assignment mark is the geometric average (the square root of the product of the squared scores)

An example of written and oral marks (boldface) vs the final mark (italic) is:

• • A student who can use AI to produce a high-quality project may score 100% (Score A), but if they are unable to explain any of the engineering concepts and receive a 0% oral grade (Score B), their final grade will still be 0%.
  • A student who produces an average project with AI (e.g. 60% – Score A) but is able to demonstrate oral mastery of the engineering principles (e.g. 100% – Score B) will still receive an above average score (60%).

The Impact and Feedback

As a result of the coupled assessments and the scoring system, students are required to have a strong understanding of the engineering principles behind their project even if they decide to use AI to generate the project content, thus ensuring that the learning process is still occurring.

Students felt that being able to use AI tools as part of their course, assignments, and end of course project was a learning multiplier, enabling them to speed up learning by getting answers quickly and accelerating laborious tasks. In their assignments they found that the AI support was most effective when they were able to identify their own initial ideas and then use the AI to refine them. Some students felt more confident in their engineering knowledge because they were able to receive immediate feedback on their ideas; however, others felt that they experienced greater impostor syndrome and would be unable to complete the work without the assistance of AI in the future.

Ongoing Work

Dr. Welland found that the addition of the oral interview was an effective solution that allowed students agency to use AI freely within the course requirements, while also ensuring that the engineering knowledge and skills were being learned by students. Administrative challenges associated with student AI ‘cheating’ were eliminated, but administrative load and human resources were increased to conduct oral interviews. For a class of 100 students with access to 4 full-time TAs, this is sustainable in the short-term but would not be feasible for larger class sizes or a course with less human resources. In the future he will explore alternatives for larger classrooms.

—

Welland, M. (2026), Unrestricted AI Access for Engineering Projects Coupled with Oral Interviews.
Licensed under Creative Commons BY-NC-SA 4.0.