Rhetorical Prompt Engineering

Bryan Kopp
Christopher McCracken
Lindsay Steiner
Louise Zamparutti

University of Wisconsin-La Crosse

This assignment introduces students to generative text technology through a scaffolded set of tasks in which they intervene in a classic professional and technical writing case study: the risk communication surrounding the Three Mile Island nuclear disaster. Students used ChatGPT to understand the case, to analyze and revise one of the memos implicated in the meltdown, to document and reflect on their revision strategies, and to develop a set of best practices for working with generative AI in technical communication. 

Learning Goals

  • Identify specific affordances and constraints of AI text generation tools for professional and technical writing    
  • Develop prompt engineering skills for professional and technical writing     
  • Advocate for professional and technical writing skills    
  • Articulate learning experiences and skills to professional audiences, such as employers 

Original Assignment Context: Multi-week unit in an upper-division introductory professional & technical writing course

Materials Needed

  • Access to ChatGPT or another AI text generation tool
  • Case study materials, including: 
    • A prompt output tracker that students can use to record their AI tool interactions. We created a Word document with a table including columns for the prompts students used in their revision efforts, the outputs they received, and their quality assessments of their outputs.
    • A “clean” (un-annotated) version of JJ Kelly’s memo.
    • Annotated versions of JJ Kelly’s memo and JF Walters’s response memo; we added annotations with additional context to these memos for students. 
    • A case study backgrounder that provides historical information about the Three Mile Island disaster, contextual information about the organization, and information about stakeholders in the case.

Time Frame: This assignment takes about three weeks (approximately 6-9 hours of instructional time) in a traditional semester.  

Overview

This assignment introduces students to generative text technology through a scaffolded set of tasks in which they intervene in a classic professional and technical writing case study: the risk communication surrounding the Three Mile Island nuclear disaster. Students used ChatGPT to understand the case, to analyze and revise one of the memos implicated in the meltdown, to document and reflect on their revision strategies, and to develop a set of best practices for working with generative AI in technical communication. 

We taught this assignment once in an upper-level “Introduction to Professional & Technical Writing” course in fall 2023. Students in this course developed an appreciation for prompt engineering as a rhetorical activity, but we found that they require some explicit guidance to think about ChatGPT as more than a mere editing tool or template-creator. 

While the writing that ChatGPT can produce is impressive, it is only as good as the prompts it is given, and effective prompts require a solid rhetorical mindset. This case study assignment is designed to promote a more rhetorical understanding of technical writing and writing technologies that addresses dynamic factors such as institutional power, multiple stakeholders with complex relationships, and human-computer interaction.

Assignment

 

Class Session Overview for Instructors

Class Session #1

  •  Prior to this class session: Divide students into teams for the duration of the project. 
  • Instructor: Introduce the case study, including some brief context about the Three Mile Island accident and share a “clean” (unannotated) version of JJ Kelly’s memo. 
  • Students: Analyze the memo. How could it be more effective? They define “effective”. Revise the memo to be more effective based on student criteria. Discuss and reflect on the process. 

Between Class Sessions #1 & #2: Students will read the backgrounder document and two annotated memos (Kelly’s memo and Walter’s response). 

Class session #2

  • Students discuss what they read prior to class and address these questions:
    • What do these memos tell you about B&W? 
    • What else do you need to know? What questions do you have that would help you understand it better? 
    • What is your communication strategy if you’re JJ Kelly?
    • What seems to be the author’s purpose?  
    • Why do you think this memo is so often held up as a failure of technical communication? 
  • Students, in teams, experiment with ChatGPT and attempt to revise Kelly’s memo. Class discussion where they explain why these revisions improve the original memo.

Class session #3

  • Overview of the case study requirements (share with students prior to class). 
  • Introduce prompt engineering techniques and do a full class discussion about how they might use ChatGPT as writers. 
  • Remaining class time for experimenting with ChatGPT to improve Kelly’s memo (students should be filling out their team’s tracker document as they use AI)

Class session #4

  • Team discussion and reflection:  
    • Review your prompts from Friday (recorded in their tracker document)
    • What prompt engineering strategies did you use?
    • Which strategies were more/less effective?
    • What do you plan to do to improve your prompt engineering next?
  • Remaining class time for experimenting with ChatGPT to improve Kelly’s memo.

Class sessions #5-9

  • Continue in-class experimenting with AI and working in teams to complete the team project portfolio deliverables. Include class time for students to discuss and reflect on their learning throughout the remainder of the class sessions.

 

Case Study Project Instructions for Students

What is this project? You will work in small teams to produce a portfolio of materials, including:   

  • Team case study debrief 
  • Revised and annotated Kelly memo (without AI assistance) 
  • Revised and annotated Kelly memo (with AI assistance) 
  • Prompt/output tracker 
  • Prompt engineering guidelines 
  • Each student will also complete a reflective assignment on their experience with this project. Please contact the authors of this assignment for details about this reflection.      

Why are we doing this project? The student learning outcomes for this project include:     

  • Identify specific affordances and constraints of AI text generation tools for professional and technical writing    
  • Develop prompt engineering skills for professional and technical writing     
  • Advocate for professional and technical writing skills    
  • Articulate your learning experiences and skills to professional audiences, such as employers 

Deliverable #1: Team case study debrief of the use of AI

  • Introduction: Briefly introduce the document and provide an overview of what you've covered in the debrief. Provide an overview of the process followed and the work completed, as a team, during this case study project.  
  • Memo Revisions: Describe the revisions that your team made to the Kelly memo. Be specific about the differences in revisions made/recommended without and with AI assistance. Your embedded comments in the two memos will explain specific revisions and/or areas for revision. You will synthesize and explain the revisions across both documents in this debrief. 
  • Prompt Engineering Guidelines Rationale: Describe the "prompt engineering guidelines" document that your team created. Explain the rationale, based on your experiences in this case study, behind both the form and content of that document. What information (content) did your team include and why? What is the reasoning for the way you wrote the guidelines? How is the document designed/structured and why? What visual design strategies did you use and why? The majority of the prompt engineering guidelines document should be written by your team; if you used any AI assistance/ideas, please describe that in this section.  
  • Formatting/length: 2-3 pages; single-spaced; 12-point readable and professional font; 1" margins on all sides; header at the top left with the name of the assignment, your team members' names, and the date submitted.
  • Note about using AI for this document: This debrief document should be written by your team; try to avoid using AI assistance for this deliverable. If you need to use it, please explain the role of AI tools in completing this debrief document. 

Deliverables #2 & #3: Revised and annotated memos

  • Include your team's revised Kelly memo in two versions:    
  • Your team's revision without AI assistance (completed during class days 1 and 2)
  • Your team's revision with AI assistance (completed after day 2) 
  • Both memos should include embedded comments with your team's commentary about the revisions made/recommended.   

Deliverable #4: Prompt/output tracker  

  • Include your team's AI prompts and output tracker as a record of your process  

Deliverable #5: Prompt engineering guidelines

  • A 1-2 page reference guide about prompt engineering and writing with AI for professional/technical writers  
  • Include strategies and examples to help "JJ Kelly" and colleagues write for their job with AI assistance; assume they are beginners at using AI text generation tools 
  • This document should help them improve their writing/communication in this case study scenario: This should be a set of guidelines for Kelly on how to prompt engineer with AI text generation tools to address the communication problems in the case study. 
  • Format this document using strategies from earlier in the semester (regarding visuals and design) 
  • This document should be primarily written by your team. If you used limited AI assistance for any aspect of the writing process or idea generation, please clearly explain that in the debrief document. 

Bryan Kopp, Chris McCracken, Lindsay Steiner, & Louise Zamparutti 

Licensed CC-BY-NC 

Acknowledgements

We drew heavily from other case studies of Three Mile Island—specifically Fennell, Miller, and Herndl’s (1991) multi-dimensional analysis of the memos. 

This assignment is part of a larger scholarship of teaching and learning project funded by a grant from the University of Wisconsin-La Crosse’s Center for Advancing Teaching and Learning. This project has been approved by the University of Wisconsin-La Crosse Institutional Review Board.