How is AI currently utilized in higher education? As of fall 2023, we are still in the early stages of figuring out what AI integration means for higher ed. In some classrooms, such as my own, we talk about what AI can and can’t do. We look at examples of output from large language models (LLMs) and critique how well it did with the responses. Next semester, my students will spend time working with prompt writing, reflecting on LLM output and reflecting on when knowledge and learning matter versus when human-AI hybrid writing makes sense. For many faculty, that last question is causing the most friction around campuses.

How the past influences the present

Instructors across the disciplines question how and where their students will use AI. Will students offload the conversion of knowledge into learning, or will students still be motivated to spend time struggling with concepts and gaining deep understanding of topics? While I see that as a valuable question, I also reflect on my own childhood. I grew up with card catalogs and encyclopedias. It was a big deal to have your own set of encyclopedias at home! We learned what we could from the books we had access to and memorized facts to regurgitate on tests. This was the epitome of knowledge and learning before the internet became widely available. Now, if my child asks me a question about anything, we can Google it on my phone from wherever we happen to be. Telling a student to go to the library to access a physical copy of something is the outlier instead of the norm.

Thinking about how information production and knowledge holding have shifted in just a few decades, I think there are lessons to be learned for higher ed. How we approach our subjects in the face of AI will both radically and also barely change. For instructors who have had the opportunity to participate in pedagogy learning groups, there may be a lot of overlap in what they already know about teaching and what changes AI might bring to the classroom. Here are some examples:

• Many people are just now coming to the idea of the flipped classroom to have students grapple more deeply with concepts before turning to AI for help. But the idea of the flipped classroom is credited to Jonathan Bergman and Aaron Sams in 2006.
• Another pedagogy model that instructors employ is the concept of transformative learning, which provides students with a disorienting dilemma of some kind and then walks them through the process of exploration and reflection to see if/how students were changed in some way through exposure to this new concept. Jack Mezirow is credited with developing this concept in the 1970s.
• The concept of Scholarship of Teaching and Learning (SoTL) is another one that instructors might benefit from thinking about. In exploring SoTL projects, instructors take time to interrogate a teaching practice of their own and see if it really works or if it works the way that they think it does. The origins of SoTL can be traced back to Ernest L. Boyer’s 1990 book Scholarship Reconsidered: Priorities of the Professoriate.
• I also find value in prioritizing process over product, which can be credited to Säljö in 1979. This theory takes away from students the stress and burden of the “three tests and a final” or “three papers and a final” model that seems to have taken hold in many places and instead focuses on how students get to the final product. There is value and learning to be gained in the journey and not just the destination. Many other pedagogical theories have stood the test of time, but these deeply impact how I approach my classrooms.

Using pedagogical theories to approach teaching with AI

So why is this gaze backward when we are talking about the future of AI in higher ed? It is a reminder both that what was old is new again and that many of us already have at least a few tools in our pedagogy toolboxes to tackle the changes that AI will bring to our approaches to teaching.

When I think about the theories presented above, I think about how I have structured and restructured my classroom over time. At one point, I completely removed reading responses because they weren’t working for me. Now they are back, but not as a product wholly unto themselves. Instead, they form the basis for our classroom discussion in the following meeting. Can students use AI to write their responses? Yes, quite easily, which is why they aren’t the only measure of understanding and why we have class discussions as well. Students turn in their reading responses before class, then get into small groups in class to discuss the reading. After that, we come together as a large group and discuss the reading. The reading responses are also low-stakes points value, taking the pressure off students to cheat. And, if you wanted to add some transformative learning questions to the reading responses, you could get even deeper into what the readings meant to the students.