I prepared these remarks for a presentation with Apurva Ashok and Amber Hoye for OER2025. The presentation was entitled “To AI or Not To AI: The Educator’s Dilemma when Creating OER Supplemental Materials,” and my role was to speak to my experience as facilitator of the Rebus Supplemental Resources Certificate program, which includes optional AI activities.
The Rebus course I’m facilitating is inviting participants to engage with AI in ways that I might have resisted if I were a student. I went into this course thinking that my human-created materials (tailor-made to the nuanced circumstances in my courses) are superior to anything generative AI could produce. I’m also opposed to the decision of Open AI (and others racing to build large language models) to scrape copyrighted content from the Internet so it can generate “new” content without permission, attribution, or citation. I have been learning about how large language models work so I can help my students navigate their use of them. But because I want to resist what I see as an ethical misstep in the development of these tools, I haven’t used them in my own writing, whether that be for my teaching or my research. I don’t want to feed the beast by putting my original human ideas into these tools. Of course, this is a little like working hard to wash and reuse plastic bags in my kitchen while most of the world is throwing away plastic water bottles (and much more). All this is to just give you a sense of where I was as I started facilitating this course.
The Most Human Element of the Course
I opened the course this summer by sharing that supplemental materials were my way into OER (often my classes don’t have a “textbook”). I’ve found that these materials–particularly the assignments–form the identity of a course. This is the place where students take the content of the course and bring it into conversation with their own experiences and goals. And I give them feedback on that work. And this human connection is what can make a course magical. Supplemental materials are also tools we can use to remove the barriers that interfere with real learning. We might create them to address a problem we’ve observed. One example from our current cohort is from a horticulture class. One of the assignments is to collect soil samples at community events. Some students are unable to attend these events in person, so the instructor, Jessica Gemella, wanted to develop an assignment so students can stay up to speed. Enter the first optional AI exercise. Participants in the course were invited to prompt generative AI to produce a list of activities that would be beneficial in their course. Though the generative AI tool this participant used didn’t “know” the particular issue this instructor was experiencing (only a human could have that insight), she took one of those ideas (a flow chart) and ran with it to envision a project that I think will be useful and fun for students and built with them. I asked if she would’ve considered a flow chart without that engagement with genAI and she said that genAI set her on this path. And I quote “I think the AI exercise helped me to pursue this project and will support me in creating and implementing it. I was originally thinking about an infographic project. The AI exercise influenced me suggesting that I was on the right track. With the option of using generative AI, I feel I may have more tools available and be able to work faster.”
Robot Teachers
Some teachers are worried about being replaced by robots. I think this is only a threat if we teach like robots. And I’ve become increasingly aware of the pressures that have been turning human teachers into robots since I started teaching (well before generative AI). One pressure that I regularly guard against is a too-rigid curriculum that leaves no room for improvisation. Another is grades, which turn real learning into a transactional exchange. And another is learning management systems that compartmentalize the many dynamic components of a course, making the human interaction hard to see. I’m regularly looking for ways to humanize the interactions I have with students, and open pedagogy and ungrading are two of the ways I try to do this work. I’m sure you can imagine that I was viewing generative AI as interfering in these human connections. That is, it makes it harder for me to do my job. An example emerged in our course that helped me see this issue of robot teaching in the context of labor in higher ed. One of our participants, an OER librarian, was told by administration that she would be co-teaching a new course on AI. Having not decided on her own to devise this course (that is, not starting from a list of her own human-generated goals), she and her co-teacher decided to start using generative AI to build the course, keeping a detailed record of the human decisions they made along the way. They will be inviting students in the course to do the same as they collaboratively create a libguide addressing the implications for generative AI for a community they’ve identified (teachers, for example). In this case, where the topic of the course is AI, transparency from instructors and students will generate authentic conversations—lots of humans talking about how they are engaging with generative AI in pursuit of new knowledge. This is one of many examples that point to generative AI–the existence of it–helping us recognize the need for human teachers. We need to highlight the humans making use of these tools every chance we get. We need to do this so we recognize that we are not replaceable by robots (and should stop trying to behave like them for the sake of objectivity or efficiency) and so that we also empower our students to claim their human contributions and use these tools rather than be guided by them.
Self-Checks
An exciting part of the Rebus course is its focus on self-checks. I’ve discussed self-checks as a resource instructors can make available for students to engage as they wish. But the second you talk about a multiple-choice question, students and instructors think about grades (and this unfortunately makes these questions more transactional and less useful for learning). I’ve been guiding participants to think carefully about how they’re framing the self-checks they create, adopt, or adapt for their courses. There are other perspectives, of course, but I think self-checks work best if learners engage with them voluntarily—not for credit (how many Duolingo users in the audience? I’m currently using it to learn Italian and I could care less about XP or my position on the leaderboard). I also discourage using self-checks as a method of surveillance (a way of checking if students are doing the reading). We need to accept that we can’t see everything students are experiencing, unless they choose to share with us. This gives learners agency. But it can be challenging to put significant work into a resource for students and then just hope that they will use it. Because we should acknowledge that self-checks are time consuming to create! One participant this summer came into the course with a robust OER that he’s been authoring over many years for an upper-level psychology course. He shared at the start of the course that he’d been experimenting with many LLMs and was “interested where AI (warts and all) might fit into the OER landscape.” He first encountered the idea of self-checks during the course and started experimenting with putting his OER into Notebook LM, asking it to generate multiple choice questions. Simultaneously, he’s working to develop an authentic assessment that has students crafting content that future students will be able to use when they take the course. He will essentially be giving students two tools for engaging with the content in his course that are very different, and that strikes me as fantastic. Even richer if he can draw their attention to the very different processes for creating both resources.
It was useful for me to see that generative AI is good at the parts of teaching that can become transactional (robots would be good at robot teaching), but that we can take that and make it a resource students can engage on their own outside of a transactional exchange. In other words, engaging with generative AI is helping me realize that there are many teaching tools that I didn’t use because they weren’t getting a good response from my students (arguably because students were viewing them as transactional and were not engaging with them in good faith). But maybe I can use them again if I think of them as resources that are there for students to engage voluntarily. I think I might even test out Notebook LM (maybe).
