OPINION: What do we do when students take the GAI shortcut?

A year ago, Senior Professor of Engineering Jens Bennedsen argued here in Omnibus that CED (Centre for Educational Development) should make generative artificial intelligence (GAI) in education a main priority. He called for “a thousand experiments” with new oral examination formats aimed at identifying cost-effective ways of assessing students’ process, given the reality that their final product may now be wholly or partially manufactured by GAI. I completely agree with him on that.

OPINION: GAI is forcing us to develop new types of examination

I teach a course in Physical Chemistry where asking students to perform calculations in the exam serves a clear pedagogical purpose. Alarm bells are probably already ringing for many readers, because ChatGPT can produce exactly those kinds of calculations to a top-grade standard in a matter of seconds. I recognise that, and it’s a real challenge. But as a full-time lecturer, I’m actually not particularly concerned about students cheating during exams. That is not the core issue. The core issue is the students’ learning outcomes.

Learning less in class

Unfortunately, we are now beginning to see evidence that appears to confirm what many lecturers, myself included, have long suspected: Not only do students perform far beyond their actual abilities in exams when they rely on GAI. They also learn less in class. That is how I interpret it when Danish upper-secondary pupils suddenly perform significantly worse in exams without GAI than they did during the school year with GAI (see DR and Politiken, both in Danish). The same seems to be the conclusion of a newly published scientific study of more than 26,000 Chinese pupils and students (although I am not the right person to assess the study's methodology). In that context, it suddenly makes sense that many students struggle with the most basic calculations in my exam, which is taken without access to GAI.

Deeply concerned on behalf of new students

I am very concerned on behalf of the new students. Today, business and industry employ fully qualified engineers who, with the help of GAI, deliver even higher levels of productivity and quality than they otherwise would. Soon, however, it will be hiring newly graduated engineers who lack the ability to assess the quality of their own (ChatGPT’s?) work. I foresee a certain degree of dissatisfaction among employers. 

I therefore want to ask: How are "the thousand experiments" going a year later? Surely we should also be conducting a thousand experiments with teaching itself. For how else can we act in time and prevent our students, the former upper-secondary pupils, from taking the GAI shortcut, bypassing the hard work of learning and leaving our programmes with artificial (intelligent) competencies that can be switched off at the flick of a switch?

Looking for inspiration and input

As lecturers, we have very little time to experiment with our teaching, so I look to CED for inspiration and input. On AU Educate, the tab Technologies mention… nothing about GAI. If I go to Practical Examples and select GAI and chatbots, I am provided with 11 examples that mainly focus on teaching and supervision of academic writing at the Faculty of Arts. A single example from Health uses ChatGPT as a reference work, while another from BSS uses the chatbot "Phil" for roughly the same purpose. But how do I transform ChatGPT from an unwelcome shortcut that sidesteps the learning process into a productive tool in my Physical Chemistry course? Unfortunately, there are no practical examples from Nat and Tech.

To be fair,... Even the major institutions abroad offer few good examples. For instance, the MIT Learning Lab only has one article on GAI in teaching that dates from after 2023. But all the more reason for a dedicated effort. I will therefore conclude with two key questions that, in my view, currently remain unanswered:

1. Is CED currently working behind the scenes on a dedicated effort to develop solutions to the above-mentioned challenges posed by GAI in teaching and exams at AU?
2. When will the senior management team act by explicitly calling for collaboration between lecturers and CED on these issues, and by providing the resources needed to make it happen?

We cannot move quickly enough on this.

GAI-declaration: In the preparation of this article, I have used GAI for... nothing. 

This text is machine translated and post-edited by Mie Skov Jeppesen.