I’m not teaching this fall, which means I have the luxury of being able to take my time to think about how I want to frame tools like ChatGPT in the classroom. This is new territory for almost all of us, and what seems reasonable to us in our particular teaching circumstance is going to vary widely. Most of my classes average at about twenty-five students, which is very different from tussling with this issue in a lecture class of two hundred. Still, I think there are some general practices that serve all of us well, no matter our situation.

• Do not default to distrust. Our classrooms are not packed wall-to-wall with students who are looking to use AI nefariously. We need to ask ourselves: does our policy about AI use communicate that we think all students will screw up, and frame the issue around the penalties that they’ll incur when they do? If so, we’re telling every student in the class that we assume they will make poor choices. In communicating our own distrust, we invite distrust back, and that’s not the basis for a good course experience for anyone.
• Be transparent. No matter what policy we land upon we should explain the pedagogical reason for our decisions (a good rule of thumb for all our teaching choices). Someone could, if they chose, ban AI completely and impose penalties for its use. If they did that, they should explain in positive terms what that policy achieves. What conditions does that create that are generative, creative, and useful for students? How can we explain that there is other meaningful support built into the course for developing writing skills, or doing research, or figuring out equations? This is good practice wherever you land on the AI spectrum—we can explain why, judging from our experiences, X policy makes sense.
• Think critically about assessment. Few people have had the time since spring to rethink all their assessment choices with AI in the mix—it’s become one more thing to juggle when educators at large are already tapped out an exhausted. So I don’t feel that any of us have to burn it all down right away. Instead, we can ask what’s one thing we can do. Can I change a particular assignment to emphasize drafting and redrafting (especially with peer feedback to make my own workload manageable)? Can I dedicate some class periods to co-working, where I have an opportunity to check in with students and answer questions as I move around the class? Can I meet with students one-on-one, or in small groups, to talk about early work (before a final product is turned in)?

 

Beyond these considerations, there are some practical, day-by-day things I’m going to fold into my classes to contextualize the use of systems like ChatGPT.

• We’re going to read about labor practices. We’ll have a class discussion about the human cost of AI, and the ramifications of asking ChatGPT a question when we take into account the lives of people who are tasked with content moderation.
• We’re going to read about environmental factors. AI needs water to generate the electricity that powers the servers, and water to cool them. The ethical considerations are enormous here when we consider global water shortages, climate change, and profit motives.
• We’re going to read about how AI actually works and what ChatGPT is and is not doing and we’re going to read about the women of color who warned us about this years ago. I’m going to follow this up with a really practical, small-scale example, and have students write a bio of themselves, or a description of how they spent their day the day before, using predictive text on their phones. Predictive text – in offering three words that often follow the word a person types into their text message—demonstrates what AI does at large scale too: estimates common combinations of words to make phrases. The bios / day descriptions students will generate will necessarily be bland, vague, or absurd (maybe all three!) and we can talk about whether they’re useful or not (including if we edit them). (If you’d like to do a deeper dive on LLMs with your students, this article is so, so helpful.)
• We’re going to read about the limitations of access to ChatGPT and the ways in which AI may stand to help – and exclude – people with disabilities unless we design with disability in mind *
• We’re going to read about editing AI with thoughtful intent
• We’re going to read data mining and privacy issues. Many students will not know what happens to the data they provide AI systems, and it’s worth learning about and taking into account. The University of Mary Washington has a wonderful learning unit about data privacy that’s available to educators, with activities we can do with our students.
• Students are going to write their own position statement on AI use. Just as I think we need to be transparent about our pedagogical choices, I think it’s incredibly useful to have students do the metacognitive work to articulate their position on AI. Will they use it?  In what ways? To achieve what ends? If they won’t use it, what has shaped that decision? Has learning about the larger ethical framework for AI helped them frame the issue differently?

I’ll make this manageable by doing a little in each class period rather than going with a one-and-done approach. It’ll take 10-15 minutes at the beginning of several classes to cover all of this, but I think it’s worth it, not least of which because AI is going to become more prevalent in our everyday choices beyond the classroom, and learning how to analyze those situations is a valuable critical-thinking skill.

All of this is, as yet, completely untested!  But I’ll write a follow-up next term about how this goes, and student reactions.

*Thank you to Ann Gagné for pointing me toward these resources and prodding me to think about how accessible ChatGPT is to users with disabilities.