The Analog Sandwich: teaching writing with & without AI
Can we bring AI into the classroom and then ask it to step outside?
How do we preserve the critical and creative art of writing and at the same time prepare students for the 21st century workforce? My one idea is to gather human experts to discuss (more on this later). My other idea is to create a semester as an Analog Sandwich.
In the Analog Sandwich (or what a yinzer like me would call an “analog hoagie”), the first section of the semester invites students to explore the extent to which AI can enhance their writing process. In the second, students completely unplug. And the third portion lets students decicide.
The Analog sandwich:*
- AI-heavy process
- Analog, an unplugged unit
- Students’ Choice
* previously I called this an AI sandwich, but the analog portion is the “meat” or the inside of the sandwich.
Unit 1: AI-heavy process
Students will be using AI in their jobs and lives even more than they are now. Much more. I’d like them to know how to use it responsibly and adeptly. In this unit, I introduce students to some fundamental concepts and then guide to use AI, although not to generate entire papers. Of course the trick of this unit is to give clear guidelines on what is acceptable vs. unacceptable usage, which is easy to say but hard/impossible to enforce.
Exercises included:
Perfect Tutor: students create a system prompt for their perfect writing tutor, an exercise invented by Jeremy Douglass. I include the work Patti Taylor has done to show the shortcomings of these systems.
Research Tools: Using LLM-driven research tools, such as Consensus, Elicit, Research Rabbit, and Ask Primo, which the library subscribed, too.
Turing Test: Students submit either text they wrote or text they generated and classmates try to guess which is which. (created with Mary Traester)
AI Energy Assessment: Using Jon Ippolito’s What Uses More, students assess how many resources they consumed. We use this little spreadsheet.
Talking to ELIZA and GALATEA: students chat with the first chatbot ELIZA or DOCTOR and also Emily Short’s brilliant interactive fiction GALATEA, which calls into question a whole host of issues involving humans and artificial counterparts.
In this section, I introduce the composition notebooks for some hand writing (or printing), including first week diagnostics, drafts of blog posts, and other in-class writing exercises. Still, much of the classroom activities engage with digital tools, including non-AI tools, such as Zoero for research and Hypothes.is for online social annotation. The paper assignment for this unit uses a lot of library database research, so students definitely need their computers.
Originally, I was going to put this second, thinking it was better to build up from the ground up (i.e., no AI → Lots of AI), but then Jeremy Douglass made the case with me that students do not have equal exposure to or experience wiht AI. Putting this unit first exposes them all to AI’s capacities before we take them all away.
This unit also includes lessons in how AI works, drawing on some key articles, so students develop a vocabulary and set of base concepts to inform their use of AI. Readings include works by Matthew Kirschenbaum; Emily Bender, Timnit Gebru, et al.; Joy Buolamwini; and Rita Raley, along with historical articles by Alan Turing and Joseph Weizenbaum (creator of ELIZA). We also are reading the MIT study of the brain development of LLM-dependent writers. Chilling!
The unit ends with some in-class hand-written reflection on their writing process for their first paper, including their AI energy use assessment, in which students use the tool developed by Jon Ippolito and his students (What Uses More) on a spreadsheet to inventory how much energy and water they used by employing AI in the production of their papers.
Unit 2: Analog unplugged Unit
In this unit, we totally unplug. We now read and annotate texts on paper. We write only in the composition notebooks. Discussions are in-person, face-to-face, eyes up. The goal is to bring us fully into the classroom space and to return students to writing on their own, unassisted. (Well, I’m happy to assist them but without digital assistance.) That also means no digital technology in the classroom experience: no laptops, no phones, maybe watches for time only. We’re reading books and photocopied articles. We are spending time being with each other and with our minds, trying exercises like Timed Thinking and Ambling Invention.
I remember early on (summer 2023) in the AI renaissance, after the coming of ChatGPT, being really skeptical of all the luddite responses — people turning to blue books and banning AI form their classrooms. That just seemed silly, impossible even. It seemed to deny the changing nature of writing. Then, I saw these composition notebooks on Dennis Jerz’s social media posts. The idea seemed ludicrous. But then I started to think about how diminished my life would be without the experience of writing — even the hard part of wrestling with a sticky set of sentences or a garbled idea or waiting (and waiting) for ideas to come. All of it. And then I realized there was just no way to ensure students got that experience if they were plugged in. AI is just everywhere in their writing environment, even if they don’t seek it out. (See co-pilot in MS Word, for example.)
In this unit, the paper is completely unplugged as well: researched, drafted, revised, and written entirely using their composition notebook. The paper assignment for this social science class asks students to interview each other about their topics and then write them up as APA-style research papers, minus the lit review that we wrote for paper #1. So even the research is happening in class.
That’s not all. Students write weekly blog posts for my class, so we are using the composition notebooks as places to draft those throughout the semester, but, even during the unplugged unit, they post these online.
Ultimately, the composition notebooks act as a a kind of true portfolio of the students’ writing — “True portfolio” meaning that it is a collection of writing over time that shows development. “True” meaning authentic. “True” meaning unassisted by machine intelligence. “True” meaning their own.
Unit 3: Student choice
I guess this is a potentially an open-faced sandwich because students will choose whether or not they go back to AI. I have more to say about this final assignment, but suffice it to say it will give students the opportunity to produce many kinds of final objects accompanied by both typed write-up and in-person reflection.
Do I think this experiment is fool proof? Nope. Ffor example, if a student has accommodations for typing, I honor those. And I am aware of the equity issues for students who typically rely on machine translation to help them with English as a second language or with spellcheck or grammar check. But I can compensate for that when assessing and grading their writing. Also, to be honest, I am not the biggest fan of in-class writing. So much of my writing benefits from my ability to copy and paste whole paragraphs. I get that. But that’s why we have the other two papers.
This is just an experiment. I’ll report back. And do let me know if try something similar.
Mark C. Marino is the Dir. of the Humanities and Critical Code Studies. He is the author of Critical Code Studies and Hallucinate This! an authoritized autobotography of ChatGPT.
