Estimated reading time: 3 minutes, 54 seconds
I am an educational developer, so I often need to help people think through how they’re going to respond to the advent of ChatGPT and the like. I myself don’t teach writing, and I think solutions for teaching writing will always be the most difficult, and their context is often quite different from courses in the disciplines. I teach a liberal arts requirement that focuses on digital literacies (including AI literacy – even before chatGPT existsed) and intercultural learning, and has lots of writing, but does not *have* to have everything written.
I know not every “solution” will work for all contexts, all courses, all levels, all learning outcomes, but some of the things I think will work include:
1. Having a lot of the written work be about reflecting on an actual real-life authnetic experience students had (like a community based learning project or interviews they did with people or a reflection on something they tried themselves) rather than be based on material they’ve read (which is easy to feed into AI). I know some students might still ask AI to write reflections for things like that, but it will appear inauthentic, I think? This semester, I’m having my students do community-based learning by teaching digital literacies to students in public schools; I’m having them do some planting of vegetables in a plot on campus (and using tech to help them do the work, AI to help them identify weeds, etc.); and meeting international students via web-based video conferencing dialog called Soliya. AI won’t be there for any of those experiences. We’ll also do a lot of games in class and I’ll ask them to reflect on the game play and connect it to readings.
2. Have some form of oral discussion around important work students have written. Conferences for writing classes, maybe, but for non-writing classes, I’d save time of “presenting” what they’ve written, and replace the “presenting” time to “discussion time” where we ask them questions about what they’ve written. If they can respond to questions about what they’ve written, then they’ve demonstrated learning (whether or not they sought help with AI)
3. To reflect on readings, I do “annotations” via Hypothes.is or similar, instead of written reflections. They select a part of the article (word, sentence, paragraph) and reflect: ask a question, connect it to their lives, connect to something they are learning elsewhere in the course, including class discussions.
4. Giving students loads of choice wherever possible to pick topics that interest them that they really want to learn about. This doesn’t AI-proof, but I think it helps a lot. Sometimes students don’t take this up and still want a shortcut, so I spend time helping them think through what they’re passionate about. Most students are passionate about something or other but don’t realize they could bring that thing into my course!
5. I am also telling my students something I said last semester. That any “generic looking writing” of the kind AI often produces (and which a B- or C student used to produce in the past) will not be acceptable any more. Any written work needs to be very specific and reflective and connected directly to their lives and to the course. Any writing that involves reflecting on written work won’t be graded on how well it summarizes the reading (it never really was, but I used to give partial credit, and now I’d give none!) but how it makes connections. Someone can still cleverly use AI to do this one, but if they manage, they’ve probably done enough reading/thinking and given it as a prompt to the AI
6. Prompt Writing Assignment: the assignment IS to write a good prompt. Prompt writing IS writing :))) I’ll make students try different prompts and keep refining them and submit a set of like 3 progressive prompts and compare the different text they produce. Has anyone done that? This one embraces AI and makes the writing assingment focused on writing for AI. It also then helps them think critically about the output of the AI and think of how to make it better… and that process is itself a learning process, I think.
Thank you to Anna Mills for encouraging me to blog this, which started as an email on a listserv of people interested in AI in writing started by Mark C. Marino.
Header image created by me on Canva. Text in the middle: Artificial Intelligence, Avoid or Embrace. Diamonds around include text that covers the 6 points above: Authentic Experience; Annotate; Oral Discussion; Student Choice; Reject Generic Writing; and finally, Prompt Writing Assignment (the only “embrace” in all of these)
This is so helpful–thank you, Maha! One thing I wonder is how to help them understand what you mean by “generic” writing vs specific and reflective writing. Might students who are more intimidated by academic writing be more likely to think they need to conform to a generic standard? I wonder if that is part of why AI detectors are more likely to flag English language learners’ prose, if those students are trying so hard to fit their words to an idea of what is conventionally correct. I guess giving them samples and discussing those in class should help.
Yes – definitely giving samples and Prompting AI in front of them and showing the results then critiquing it for the genericness? But reprompting multiple times can produce better results … but it also means we are doing more thinking and writing ourselves. What do you think?
Yeah, I think sometimes it produces writing we wouldn’t consider generic. And I worry that telling students not to write something generic could create anxiety since that could be defined so many ways. I don’t want students to feel its on them to figure out what kind of writing their instructor might perceive as AI and avoid it. I know you’re not trying to send that message, and it’s good to write specifics anyway, but it’s a concern that they’ll worry. They might also just prompt ChatGPT to be specific, no?
So my sense is that, in any case, teachers need to be clear with students what they mean by quality work. I don’t know if rubrics are ever clear to students, but samples to discuss, maybe? I do think that if we’re not teaching writing and someone prompts AI to write something more specific, the prompt itself probably includes the specifics, and so it might be OK? If it’s something that connects personal reflection to something else? I have recently noticed big differences between how well GPT4 does this vs free ChatGPT’s latest version