Eighty students stare out at me on the first day of the semester, waiting for me to begin lecturing so they can disappear into their own thoughts, their laptops, their surreptitious texting with friends. It is the perfect embodiment of how large classes have fuelled the depersonalisation of higher education.
That’s exactly why I, in the first 10 minutes of the class, introduce them to their “teaching assistant” by telling them to log into a frontier AI model such as ChatGPT, Claude or Gemini. I share with them a prompt to cut and paste into the AI – “I don’t really understand why I am taking this general education course. My professor says it has something to do with learning how to be a critical thinker. What does that even mean?” – and tell them to have a quick “conversation” of at least five back-and-forth responses.
- Every AI learning persona needs an origin story
- How can we teach AI literacy skills?
- Bringing GenAI into the university classroom
My prompts are actually much more detailed than that, including specific context of the issue, directions to the AI on how to have a factual and supportive conversation with the student, and the opportunity for students to personalise the prompt based on their academic major, professional goals and level of understanding. I create a unique one for every class topic. This is usually called “prompt engineering” or “strategic prompting”; I also employ “meta-prompting”, where the AI basically helps me build each prompt. Long story short: I have created strong guard rails and guideposts for the ensuing conversations with AI.
I give my students a few minutes and then I model it myself in voice mode, allowing the AI’s voice to boom out from the lecture hall’s speakers (as I connect my phone to the lecture hall’s Air Play system). I interrupt the AI, ask it for examples, make it explain something as if I were a third-grade student, then as if I was a postgraduate student, then have it summarise our conversation, and finally turn this summary into a funny four-line poem.
I don’t mean to be too dramatic (although maybe I do) but in that moment everything changes. I explain to my students that from now on, I will expect them to always have their AI tutor open in class, asking it questions about anything and everything they may not understand in my lecture.
I also explain that they will have weekly “AI Labs”, where they will be required to have even more in-depth conversations, and that I have created prompts to make the AI push them to think hard about why they believe what they do (which is basically a form of Socratic questioning). Part of the prompt goes like this: “It is important that you push me on my perspectives, as I learn best by being questioned and pushed to think deeper about my answers. If I give vague answers, ask follow-up questions that invite examples or evidence.”
All of this, to put it mildly, surprises my students. “My friends and I only used AI for cheating,” one student wrote in the anonymous end-of-semester evaluations, “so I had no idea it could be used as a study partner.”
“At first,” another student wrote, “I will admit that I found the AI Lab a little funny and wasn’t really sure how to interact with it, but I love the idea of it being a teaching assistant now.”
Probably my favourite (although dozens of other students wrote some variation of this) was: “The AI Labs are extremely interesting and I’ve never done anything like them before. I like that I can talk to AI about the content and be asked deeper questions that I wouldn’t think of. They have helped me think deeper and learn new things about myself. I feel like they have also helped me be a better writer.”
I want to be clear that getting to this point wasn’t easy. It has taken me three years to figure out how to use AI as a tutor, how to stop my students from cheating and how to think through my own role as a professor. Many times, I was ready to give up.
But today, with AI as my teaching assistant, I actually think I am a better professor than I have ever been.
I say that because my job is to teach my students about complex and contested issues in our schools and society. And that’s hard. It’s hard because higher education is better at sorting students than teaching them. It’s hard because critical thinking itself is hard; as one recent meta-analysis put it, thinking is “unpleasant”. It’s hard because, in that large lecture hall, my students all too often feel unseen and unheard.
But I want to suggest that today I can give each and every student in that large lecture hall the chance to think through and talk through and understand everything I teach. In other words, when used in the right way, AI can help us rethink the personalisation of higher education.
Dan Sarofian-Butin is professor in the department of education and community studies at Merrimack College.
If you would like advice and insight from academics and university staff delivered direct to your inbox each week, sign up for the Campus newsletter.
comment