The Paranoid Memorandum

A Generative AI Exercise for Professional Communication

Jason Crider
Texas A&M University

This classroom activity engages students in an undergraduate technical and professional writing course in the critical evaluation of workplace communication alongside the specter of AI writing platforms. In small groups, students draft workplace memorandums according to prompts featuring an imaginary scenario. At random, each group is told the degree to which they can, cannot, or must use AI to author their memo. Finally, we take turns critically evaluating and revising each group’s memorandum.

Learning Goals: 

  • Consider how ChatGPT and other AI assistants now “haunt” the space of writing and how as writers we might think more strategically about how to use these tools, read with these tools in mind, and write in a way that still feels authentic even within a formulaic genre like a memorandum
  • Offer a starting framework for leveraging AI skepticism into productive learning outcomes moving forward in the writing classroom

Original Assignment Context: Undergraduate technical writing class

Materials Needed: An accessible AI text generation program (i.e. ChatGPT),  genre exemplars

Time Frame: ~1 class session


Like many of my colleagues, I spent the winter break prior to the spring 2023 semester trying to figure out how, where, and if ChatGPT fit into my writing pedagogy. I had initially hoped to experiment with using it in my classical rhetoric graduate seminar, but it proved to not really be useful as an assignment or activity for my students. As a tool that feels particularly suited for genre mimicry, I thought for sure I could figure out a way to use it to generate AI Socratic dialogues or autoprogymnasmata,1 and yet my experiments repeatedly yielded unsatisfying results. My only takeaway from these is that ChatGPT does not (yet) appear capable of engaging in dialectic. How could it? If, as James L. Kastely suggests, the role of the rhetor in a good-faith dialectical exchange should be to offer oneself up for refutation,2 what would it take for an AI assistant to actually be capable of engaging in one?

Dialectics aside, I knew that I had to embrace ChatGPT for my undergraduate technical writing class. I ran a number of my previously assigned informal weekly writing response prompts into it and was unsurprised to find that the program could adequately complete many of them. And of course, much of technical writing, at least at the introductory, undergraduate level, deals in formulaic and often templated genres of writing. Technical writing seems to me the exact type of writing discipline that can not only greatly benefit from tools like this, but also drive the kinds of innovation that could promote ethical and generative frameworks for them. The central questions then for myself and for my students remains exactly the same: What does writing produce? What am I expecting my students to learn by writing and thinking about writing? And how can I tell if they are actually doing that learning? 

I started with a syllabus policy:

Artificial Intelligence Policy*

Students are permitted to use AI assistants, such as ChatGPT, to assist in their writing process in this course. However, there are certain guidelines that must be followed to ensure the integrity of the student's work.

  1. An explanation of prompts must be submitted alongside the assignment, as well as a brief summary of how they were helpful in drafting the assignment. 
  2. Students are responsible for fact-checking all information generated by the AI assistant. Any inaccuracies found in the final submission will be considered a violation of academic integrity. Please note that although these programs are very good at creating answers that sound authoritative, there is nothing on the backend that performs any fact checking.
  3. AI assistants should be used as a tool to improve the student's writing skills, not to cheat on assignments. Submitting work generated entirely by an AI assistant will be considered plagiarism.
  4. Students who violate these guidelines will be subject to the university's academic integrity policy, which may include, but is not limited to, failing the assignment and/or disciplinary action.
  5. Students are encouraged to discuss with the instructor if they have any questions or concerns about using AI assistants.3

*This policy was generated by ChatGPT and revised for clarity.4

On day one of the class, I spent close to thirty minutes fielding questions about this policy (often with follow-up questions of my own), leading to probably the single best first-day-of-class discussion about writing I have ever had with my students. I think the most generative question I asked them was, essentially, “why am I asking you to write?” This opened up the discussion into one about writing as an ongoing cognitive process. I made it clear to them that I have already read plenty of “perfect” essays (derivatively speaking; in terms of grading). In other words, I am not grading a final product; I am grading your ability to think critically and rhetorically with and about language. As this was the honors section of the class, I also invited them to help me experiment with ChatGPT throughout the semester and told them I needed their help to invent practices and policies for using tools like this in future technical writing courses.

The Assignment

The Prep

For this in-class activity, students will break up into small groups of 3-5 to collaborate on a fictional workplace memorandum. Ideally you would have one class period devoted to introducing the various genres and conventions of workplace communication, followed by two class periods dedicated to collaboration and workshopping. This activity could work well in both synchronous and asynchronous teaching environments, as well as in either in-person or online modalities (and in fact, may work better with the affordances of online breakout rooms). 

The main preparation consists of two elements. First, perhaps obviously, is to model some uses of ChatGPT for the class. For this professional writing unit, I often use television sitcoms set in a workplace to establish a fictional rhetorical situation.5 As a class, practice analyzing the rhetorical ecology of a particular scene or episode and then use ChatGPT to draft an email (or series of emails) as one of the characters that addresses a specific exigency within the episode. When I first did this, many of the students who had not yet interacted with AI writing assistants were surprised and impressed with the output. But as we took a closer look, the cracks started to appear. One example: “This email is formatted well and serves its purpose, but that’s not how Michael Scott would talk to Toby.” When I asked how Michael Scott talked, we quickly ended up in a dialogue about how to best characterize his attitude and delivery. In order to get that to translate through ChatGPT meant carefully and methodically experimenting with very specific word usage. We were talking about what it means to write!

The second element of preparation consists of deconstructing the common workplace documents like the memorandum, report, letter, white paper, etc. For background on these, I use the open educational resource (OER) technical and professional writing textbook, Howdy or Hello?, developed by colleagues in my department.6 We also look at a number of (in)famous and/or fictional memorandums, such as humorist David Thorne’s “McMemo,”7 to practice analyzing how memos address multiple audiences, how they circulate, and how they play into and against the conventions of their genre.8 I particularly like an example workplace incident report found in Richard Johnson-Sheehan’s Technical Communication Today (115-16).9 When students are tasked with locating instances of passive voice or nominalizations within this sample text, a secondary reading emerges in which it becomes clear that the fictional ChemConcepts, LLC is not only gaslighting its employees about workplace safety protocols, but also recklessly setting themselves up for litigation (as one student pointed out, some likely tertiary readers of this document would be the employees’ union representatives and/or lawyers). We also explore more granular elements of workplace writing, such as the rhetorics of grammar (for example: how do we assess the variations in tone in these sign-offs: “Thanks,” “Thank you,” Thanks.” and “Thanks!”?).10

The Writing

After establishing sufficient background knowledge, the students can break out into their writing groups. Each group should have at least one “prompt giver,” someone who is comfortable interfacing with ChatGPT, and ideally everyone has a laptop and can collaborate on a shared document. On the board, list out various prompts for fictional memorandum and letter writing and assign one to each group. For example, one of my prompts was: “Using TAMU letterhead, write a transmittal memo from transportation services addressed to all students, faculty, and staff that informs them of all of the policies related to the new cross-campus zipline.” Another: “As a public relations representative for the university, write a formal refusal letter to Quentin Tarantino’s request to film a particularly violent scene from his new movie at a highly recognizable location on campus. He is the keynote speaker for an upcoming film studies colloquium on campus, so try your best not to alienate him.” I recommend choosing prompts like this that are just a little bit off or that require specific, local knowledge that the students all share, as it pushes students to author more dynamic, critical prompts. 

Next, explain that each group will receive a notecard with secret instructions. They will receive one of the following cards:

  • Create your document using only ChatGPT. You may style the document after the fact using what formatting, letterhead, embedded links you see fit, but you must only use text that was directly generated by ChatGPT.
  • Draft your document using only ChatGPT. Once you find a structure or outline that you like, you may do line edits as you see fit.
  • Draft your document without ChatGPT. Once you have a completed draft, use ChatGPT to edit, revise, or alter your draft in some way.
  • Do not use ChatGPT at all.

After all drafts have been completed, read and analyze them as a class and task them with trying to determine to what extent AI was used on each document. Like in a typical writing workshop, the students in the group that wrote whichever given document is being analyzed are not allowed to speak on behalf of it or their rhetorical decisions. When discussing a document, students must justify their rationale for why they think it belongs to a certain category. A lot of these answers will be something along the lines of “it just feels too robotic,” “I don’t know, I can just tell,” “it feels too polished,” etc., which are all perfect opportunities to ask them to point to specific instances in the document that they believe are producing this “gut” feeling. After discussing each document, hold a vote for each one and write the tallies for each category on the board. 

The Trick

Every group received a “Do not use ChatGPT” card. 

The Takeaway 

I was shocked by the results. First, out of the 22 students in the class, only two suspected my “trick” (fortunately, they were kind and savvy enough to keep it to themselves until right before the reveal). Second, there were less than 10 total votes cast for “Did not use ChatGPT at all” across all five of the documents. Perhaps less surprisingly, there were only 10 total votes for “Only used ChatGPT.” That said, I do think these results would have been the same no matter how I had distributed the AI instruction cards, as the suspicion of AI now looms over all writing. It is also worth mentioning that I only “tricked” them because it felt like the fairest way to distribute the labor of the activity. We ended up having a long and nuanced discussion about how ChatGPT and other AI assistants now “haunt” the space of writing and how as writers we might think more strategically about how to use these tools, read with these tools in mind, and write in a way that still feels authentic even within a formulaic genre like a memorandum. Due to the success of this activity, I plan to run a modified version of it as part of a two-week resumé and cover letter unit. Students will read Beatrice Nolan’s Business Insider article, “I asked ChatGPT to write my cover letters,”11 alongside some other short, supplementary readings on the genre conventions of application documents.

One of the strengths of this activity is that it is platform agnostic and can be adapted for countless other writing exercises. It helps center both the process and product of writing in a way that I think really resonates with students in this specific digital moment. While many of us remain skeptical and somewhat pessimistic about many of the implications of tools like ChatGPT, this activity might offer a starting framework for leveraging that skepticism into productive learning outcomes moving forward in the writing classroom.12

1. Thanks to Andrew Pilsch for his help brainstorming AI rhetoric ideas

2.  James L. Kastely, “In Defense of Plato’s Gorgias

3. I plan to update these policies dramatically as generative AI continues to evolve. I think it will be crucial for writing teachers to move away from punitive models of assessment and towards models that put the impetus on students being responsible for the output of such tools.

4. Thanks to Lisa Messeri for the idea. See:

5. Thanks to Raúl Sánchez for this idea. I have found Parks and Rec, Superstore, and The Office to be particularly useful for this.

6. Matt McKinney, Kalani Pattison, and Sarah LeMire, Howdy or Hello?: Technical and Business Communications - 2nd Edition.  

7. See:

8. I use excerpts from Gary Olson’s Style in Technical Writing, Joseph M. Williams and Joseph Bizup’s Style: Lessons in Clarity and Grace, and the previously mentioned Howdy or Hello?. We also look at excerpts from things like The IBM Style Guide and MailChimp’s Content Style Guide.

9. Richard Johnson-Sheehan, Technical Communication Today - Fifth Edition

10. Thanks to Arley McNeney for the idea. See:

11. Beatrice Nolan, “I asked ChatGPT to write my cover letters. 2 hiring managers said they would have given me an interview but the letters lacked personality.”

12. Special thanks to Natalie Goodman for all of her ongoing help, inspiration, and generosity in helping me think through AI, writing, and my teaching. And thanks to Mike Frazier and Jentery Sayers for their invaluable feedback on an earlier draft of this piece.