I have no sympathy for a professor who worries that students could use a chatbot to cheat on an exam. Chatbots and their relatives can be helpful for student assessment if you want them to be.
Is an assessment going to consist of a grade or constructive feedback? It ought to be a combination of both.
As far as a grade is concerned, there is inherently an incentive for the student to try to obtain as high a grade as possible without earning it. There will be no foolproof way to stop a determined cheater. The best one can hope for, in a world with or without large language models, is to make cheating sufficiently difficult to get away with that students opt not to cheat.
Constructive feedback is much closer to a win-win. The professor should want to provide constructive feedback, and the student should appreciate receiving it.
Use AI for assessment
If I were a professor, I would be using AI to provide grades and constructive feedback. This could be an ongoing process. It need not wait for a final exam.
It took me only a few hours to train this essay grader for ChatGPT (and now you don’t need a paid account with openAI to try it). If I were teaching a course, I could take more time and produce an even better grader. I could also review the grades and feedback before forwarding them to students.
Now that the chatbots have gone “multi-modal,” there are ways to discourage cheating. I am sure that a determined student could find a way to defeat any system, but I repeat that the goal is to make it costly to attempt to do so.
The chatbot could give an oral exam, so that the student has to think on his feet. The exam could be in the form of a discussion. The chatbot would ask follow-up questions, probe deeper, and see how well the student can defend his answers and handle objections.
You could also imagine an AI that can proctor a student taking a written exam. The AI could have a camera positioned to look over the student’s shoulder as the student writes or types his answers.
The AI as mentor
If I were a professor, what I would really like is for the chatbot to be trained to convey the subject as I would. It would be able to tutor a student the way that I would. In response to the student’s ideas and questions, it would suggest articles for the student to read. It would give the student issues to think about. It would have regular discussions with the student.
After several weeks of working with a student, the chatbot would be able to assess the student on many levels. It could provide prospective employers with detailed descriptions of the student’s strengths and weaknesses.
The more that I think about it, the less that I think AI represents a threat to the quality of education. Instead, it seems to me that it represents a tremendous opportunity.
Don’t forget your null hypothesis! In the 1990s “learning sciences” integrated computers and teaching/assessment, with the ideal product something game like that helped students learn. I don’t think it made much of an impact.
AI doesn't pose a threat to education; it poses a threat to educators.