At today's department meeting I asked about the progress of the faculty's policy/advice on generative AI use in writing MSc theses. It was announced to us before the summer that it was forthcoming. I was told it was unlikely to arrive and that, if it did, it wouldn't be anything different than how academia treats AI written journal articles. I was surprised to hear this because I do not think an educator assessing an MSc thesis is the same thing as a peer reviewer assessing a journal article.

The journal article peer review system's primary purpose is to ensure valid and correct science is published with a stamp of community approval. There are no pedagogical goals in the assessment of a journal article, it is presumed the authors learned to do science and write a paper elsewhere. So, in some sense, it may not really matter who (or what) wrote the journal article, so long as it has correct and valid scientific information present in it.

An engineering Masters of Science thesis may look a lot like a journal article but educators not only strive to ensure valid and correct science was performed but we also use it to ensure adequate learning by the student. Reading a thesis that is truly written by a student lets you see into their mind, their thinking, and their reasoning as well as whether they can produce both the described scientific work and communicate about it with their own words. A human written thesis has long been an excellent product to know if a student has obtained specific knowledge and can perform specific skills (which include the ability to write).

But if a student uses generative AI to produce the content of the thesis, it becomes impossible to use the written words as an assessment of what the student knows and can do. Additionally, if the student also used generative AI to do the described science, we also no longer know if the student has acquired the skills described in the learning objectives. Educators around the world all know this and there is widespread panic from those that see the elimination of our ability to assess a student's work as the profession has done for hundreds and hundreds of years.

Our faculty's MSc assessment process puts great weight on the thesis and a student simply cannot graduate and become a certified engineer without one. But we can no longer assess a large portion of the MSc learning objectives via the thesis if we cannot rule out generative AI use.

I currently think of two ways out of this pickle:

1) The faculty changes the learning objectives to simply assess whether the student can successfully make use of AI to produce valid engineering.

or

2) The faculty changes the assessment to something where generative AI cannot be leveraged.

I think 1) is the wrong move, for the same reason that calculus courses do not allow the use of computers when assessing whether you understand integration and can solve an integral. It is a disservice and dangerous to society to brand people as engineers who lack the long expected knowledge and skills. The student should have to exercise their brain in the same way we always have had to. That leaves 2) as a path forward, but what is the assignment and assessment that has some equivalency to a thesis project where generative AI cannot be used? The only ones I can quickly think of require much more educator labor resources.

The faculty and university need to address this issue more critically. Society expects we produce engineers that possess certain knowledge and skills regardless if the AI crutch is present or not.