I have spent a lot of time with Judge Hammer over the past two years handling cases of inappropriate and unauthorized AI use (and I’ll have you know I am very good at detecting AI). I had a strict no-AI policy in my classes, and that included requiring students in my intro to sociology class to initial a statement indicating they understood that, if I suspected generative AI use on any assignment, I would submit it as a potential honor code violation (this did work to deter the more obvious attempts at cheating . . . I think). In general, I would describe my previous self as full of rage about AI, and people in Boyle Hall probably started avoiding me because they knew I’d have a new AI complaint to share.
Through this rage, though, I realized that my approach was not only unsustainable, but it was unrealistic and probably also unfair. Policing AI was not a good use of my time, and I understood that many students were using it as a learning tool and not simply to write their papers for them. Thus, I decided to challenge myself by taking a new approach to AI in my upper-level elective on religion and inequality in the U.S., which included a mix of sociology majors/minors and honors students.
First, I created an AI policy and explained that I allowed limited AI use on all assignments (unless otherwise indicated) and that students simply needed to include an “AI statement” that explained how they used AI in their work. Looking back, my policy was too long and detailed, but I was nervous and wanted to give ample examples of appropriate ways to use AI. Students had to develop their ideas and do the actual heart of the writing themselves, but I encouraged them to explore AI as a thought partner, as a friendly editor, etc.
This approach of allowing AI, with an appropriate explanation, worked well, especially on the reading responses that they could submit almost every class period. About 30% of the 25-person class used AI consistently, maybe 25% used AI occasionally, and the remaining 45% never used AI. I learned a lot about the different ways that students were thoughtfully, responsibly, and ethically using AI. For example, one student would have ChatGPT make editing suggestions while not changing any of the original writing; another student would always ask ChatGPT to compare their draft to the rubric, since this was their first sociology class; one student would input their original reading response and ask for help brainstorming a sociological question. This approach left me confident that I was always reading student work, and I didn’t have to pull my hair out wondering if a grammatically perfect response came from a student – it did, but with authorized editing help from AI.
Second, I designed an assignment with the help of AI that then required students to use AI. Getting feedback from ChatGPT genuinely made the assignment better, because ChatGPT identified places where I could clarify my expectations, generated additional examples for me, and brainstormed an initial rubric (which was terrible and required substantial modifications). I included an AI statement of my own to explain my use of AI, because that was the policy for the class, but I felt uncomfortable “admitting” how ChatGPT helped. However, I would want to maintain this transparency moving forward, and I think having an AI statement for the whole class – and holding myself to that same policy – will be beneficial.
Third, I gave students the aforementioned assignment that asked them to submit their analytic memo (basically their first attempt at making sense of original interview data) to GenAI and then to use at least five prompts to solicit feedback. I gave them suggestions based on my own experimentation with ChatGPT, which many people used, and then I asked them to compare my 100% human generated feedback on that same analytic memo to the AI output. Finally, they answered some reflective questions about their use of AI. I was really happy with how this assignment went . . . and pleasantly surprised by how good AI was at giving concrete suggestions and examples to push students’ qualitative analysis further. Upon reflection, I think students were rightly suspicious of AI’s suggestions, because they do not have the expertise in qualitative research methods that I do to recognize AI was actually right. I found this to be both encouraging and discouraging; encouraging in that students didn’t blindly do what AI said, and discouraging in that they were missing out on excellent advice. Many did say, however, that they had not used AI specifically in this way and that they could envision getting higher-level feedback from AI on writing assignments in other classes, if allowed.
I did not do enough to teach about how to effectively use AI, to discuss the environmental harms and ethical concerns of AI, etc. When you’re teaching a class that’s in your research wheelhouse, it’s hard to give up a minute of classroom time to anything but that subject. However, I think moving forward I will include more direct conversations about AI, which thankfully will be easy enough to connect to class material.
AI Statement: I did not use AI in any way to assist with writing this blog post 🙂
