Heidi A. McKee
Miami University
In this project, via a series of scaffolded assignments, students selected and read medical journal articles and then drafted and revised research summaries for lay audiences, exploring, analyzing, and integrating the use of AI writing systems throughout the process. This assignment is adaptable to a variety of undergraduate and graduate courses.
Learning Goals:
Original Assignment Context: mid-level undergraduate professional writing for healthcare course
Materials Needed: An AI text generation program (e.g., YouWrite, ChatGPT), library access to peer-reviewed research
Time Frame: ~2-3 weeks
Introduction & Project Explanation
In spring 2023, I taught English 310: Professional Writing for Healthcare, a synchronous online course to students at Miami University. Students in the course were from many majors: Biochemistry, Biology, Kinesiology, Nursing, Professional Writing, Speech Pathology, and Strategic Communication. Throughout their work on various projects, students used large-language model artificial intelligence writing systems (mostly ChatGPT) at various stages of the writing process. In this assignment description, I focus on the first project for the class, because it also includes a lot of foundational steps we did as a class to explore and reflect on AI writing systems and human-machine teaming in writing.
For the first project, students first learned about the genres and rhetorical contexts for lay research summaries, reading and analyzing sample blog posts, press releases, and newsletter articles. Next they each selected a peer-reviewed research article in an area of healthcare they were interested in, and wrote a human-only summary of the article in a lay summary genre of their choosing for an audience of their choosing. They also peer responded to their human-written drafts. Then they used several AI writing systems available for free in January 2023 (QuillBot Summarizer, You, ChatGPT—3.5 version) for a variety of things, including to summarize the article, generate a summary from their notes, and rewrite their summaries. Students then wrote a mid-process memo reflecting on their writing and AI’s writing. And, finally, they drew from their drafts and AI’s drafts to write a final research summary, providing an annotated copy where they highlighted and explained what words/phrases/sentences and ideas came from AI in their final draft. I include all of the assignment prompts below.
The goals for the project were to provide opportunities for students to learn about writing research summaries for lay audiences and to learn about how they might approach writing with and, via prompting, for AI systems. This project worked really well, and its general steps of (1) writing human-only, (2) using AI writing systems, (3) reflecting, (4) revising, and (5) annotating a final draft can be applied to any writing assignment. For other projects in the class, such as social media campaigns for healthcare organizations, students started with AI-created drafts and then added in human approaches to revision, but for this first project we started with humans because I wanted students to really dive into their research articles and know them as best as possible. Given the errors AI can make in writing, particularly about complex research articles with a lot of quantitative and/or qualitative data, it was essential for students to know their articles well so they could discern errors and misstatements and effectively analyze AI’s contributions to the process.
I could say human-only pretty confidently because in January 2023 most students had never used AI prior to the class. This first project in the class was for most students the first time they had used large language model AI writing systems, such as ChatGPT.. In the future, this will not be the case, so what I will do is engage students even more extensively in studying the limitations and regressive bias of large language models, and I will continue to emphasize the importance of knowing their material so as to collaborate with AI effectively and ethically.
Some of the students were nervous at first about new technology usage, but after assignment 3 below (the day we explored some of the AI systems available, including ChatGPT), they realized that these systems were easy to use and they were less nervous. In their mid-process reflections, students noted many of the issues that have been identified with AI, how it is able to generate text quickly, including how a machine reads (through Natural Language Understanding and Natural Language Processing) other texts and summarizes them. Some students gained insights into their articles through what AI generated, causing them to realize they had omitted an important part of the article in their own human-draft summaries. For example, one student in her summary had focused extensively on the medical impacts of a particular disease and treatment, but had not discussed the social impacts on the patients. Reading the AI-generated summary (where ChatGPT had discussed social impacts) and comparing it to her own, she realized she needed to center patients and the social impacts on patients more in her final summary.
Through asking AI to write summaries of the article and asking AI to revise their own writing, students also gained ideas for specific words, phrases, and even sentences to include in their revised drafts. Many students appreciated how AI, especially ChatGPT, was able to take some of the complex medical jargon in the article or convoluted sentences in their own prose, and revise them to make them clearer and more tailored to lay audiences. Students became adept at being AI prompters, asking ChatGPT to do things like (and I paraphrase), “Please rewrite the following so that a 10-year-old could understand it.” Students also appreciated using AI for final copyediting: “Please correct all grammar, punctuation, and spelling mistakes in this and provide explanations for your corrections.”
But amid the benefits of AI, students also recognized the problems AI had in terms of mistakes in summarizing or mistakes in tone, like using some language that would be insensitive in relation to the population being studied in the original journal article. Students also recognized that much of the AI prose was more bland, lacking their own voice and style. By the end they almost all had integrated AI writing into their final summaries in strategic and select ways, a few sentences here, some words or phrasing there. In their mid-process and final memo reflections, students explained why they chose to include what they did for AI and many of them noted the need, always, for human decision-making and human agency in the writing process.
I bold and emphasize that last sentence, because, like my students, I want to emphasize the necessity and centrality of human agency in human-machine collaborations. Students (and all writers) need to recognize that they may integrate AI throughout the writing process but just because AI writes quickly and grammatically correct does not mean its writing should always be used. The mid-process reflection was important for helping students take the time to dive in and analyze what AI had produced and what they produced and helped set them up for revising to a final draft. The annotations on the final draft worked well too and were important for me and students to know what was from AI and how students adapted and revised to integrate AI into their writing. But what I realize is that my prompts did not do enough to forefront the student’s decision-making power so below in brackets [ ] I note what I would add to the prompts.
One challenge early on probably won’t be a challenge anymore as AI systems proliferate, but I mention it here, perhaps more as historical record-keeping. For the day I introduced AI to students and we discussed AI writing systems, one challenge was that the best available system at the time (ChatGPT) was often not available consistently—users would get blocked if too many people in the world were on the system at once (and this was before the paid, guaranteed-access pro version was available). The week prior to the showcase I asked students if they felt comfortable doing so, to get ChatGPT accounts, and luckily the day of the showcase one student did have access, so she shared her screen and we all gave her prompts to enter so we could see ChatGPT in action. Another challenge was that some of the free AI systems have a character/word limit on uploads for summarizing text, but students got around this by entering their research summaries a few paragraphs at a time.
After this assignment, students continued to use AI for the other projects in the class, including revising the text of a badly written state of Ohio public health brochure, creating a social media campaign for a health organization, and conducting and reporting (orally and in writing) on an independent research project in healthcare communications (e.g., communicating about colon cancer to African American communities, considering how to convey risk communication in genetic counseling, increasing LGBTQ+ social media outreach in hospital communications). For some projects, some students who were bilingual speakers of Spanish, used ChatGPT to help translate English text to Spanish so as to create bilingual communication materials. As of this writing (May 2023) the course is still in session, but students overall and I have really enjoyed working with ChatGPT (and now, GPT4). We are definitely encountering some of the biases—particularly in its writing around sensitive health issues and writing to underrepresented communities—but overall we all are finding the use of large language model AI systems beneficial in various ways throughout the writing process. By scaffolding throughout assignments and activities for reflection and critical analysis, students are able to engage deeply with their own writing, their own writing processes, and the dynamics of human-machine collaboration in writing.
In the course we also have many guest speakers from industry, including some professionals who now have company-paid subscriptions to ChatGPT because they are already using the system extensively in their work in healthcare communications. So in ENG 310 with the many opportunities to critically engage and write with AI systems, students are gaining valuable experience and critical literacies to prepare them for the fast-changing dynamics of professional writing in the years ahead.
1. Overview of Research Summary Assignment
In this project, we will examine a significant area of healthcare writing—writing about research for lay audiences.
Medical and scientific research articles are complicated documents filled with specialized, jargon-filled language that can be very hard for non-medical specialists to understand. Thus, an important area of healthcare communication is to make important health information and research more accessible to everyday people, often in the genre of the press release or the newsletter or blog posting.
For this project, you will find a peer-reviewed research article in an area of medicine and health that interests you and you will summarize it in various collaborations with artificial intelligence writing systems.
You will first read and write (on your own, no AI) a summary (350-550 words) of the article in a press release or blog/newsletter format for a lay audience that you identify (e.g., pregnant women reading Pregnancy magazine; parents who subscribe to blog about asthma). We will have a peer respond on that draft in class.
Then you will explore and experiment with various AI systems (explained in assignments below), analyzing your own and AI’s writing. Working with AI and your writing you will revise your summary based on feedback from peers and ideas/language gained from the AI summaries to produce a finished summary.
You will also complete a reflection on the process of writing with AI, considering your experiences as a writer and as a writer collaborating with various AI systems in the drafting process and what ethical and composing issues you could see arising with these systems, and why.
[It’s important to remember that you are the writer with the agency and final decision-making. AI writing systems such as ChatGPT are impressive in how quickly and (mostly) correctly they can generate writing, but that does not make what they write the best option. You have a unique voice and critical insight as a person and rhetorician familiar with the audience, purpose, context for which you’re writing, so definitely consider carefully what you wish to include in your final write-up.]
2. Readings & Samples on Writing about Research for the Public
[I do not include this assignment here just to get to AI more quickly, but if anyone would like the many resources I’ve gathered on this, please let me know. The key point is that students spent time learning about and reading about two (of the many) genres of lay summaries: newsletter/blog entries & press releases.]
3. Find, Read, & Take Notes on Research Article
For this assignment, you need to find, read, and take notes on a research article in any peer reviewed journal that relates to health and medicine. Then craft a scenario where you are writing a summary (either in the form of a press release or a newsletter/blog post) for a particular publication and for a particular audience. Your summary should be ~350-550 words written single-spaced.
Find and Save the Article: In the Miami databases and/or on the web via search look for peer reviewed journals in areas you're interested in nutrition, public health, psychology, speech pathology, cancer, etc. Find a research-based article (the full article, not a summary of the study) and for this assignment, please post the pdf of the article you picked to the project Google folder so peers and I when reading your drafts will have the article easily available for review.
Read and Take Notes on the Article: Please take notes on the article and upload those as well to Google.
Also, very important: Please make sure you can copy and paste text from your PDF. If you cannot highlight text and copy it into a Google or Word file, then it will not work for machine reading AI systems we will use next week. In order to use these systems, the PDF file of your research article must have its text recognized so you can either cut-n-paste from it or you can upload it and the machine can “read” the text. Many pdfs are set to default as images so the individual words in them, although we can see them, cannot actually be read by text-reading systems. You can run optical character recognition (OCR) on your pdf so that text in the pdf is readable by machines. But some pdfs are locked down completely with password protection. So be sure your PDF will be machine-readable.
4. Draft a Summary & Peer Respond
Please write a rough draft posting in Google by [date/time] and please read and comment on a [assigned] peer’s draft by [date/time]. Please identify at the top of your document the audience and context and genre for your draft. Be sure the pdf of your article is available for peers to read. As long as you have a robust draft and provide thoughtful and detailed feedback to a peer that shows you have read their article and read their summary well you will receive full credit on this assignment.
5. Class Activity: AI Writing Systems [full class period]
Artificial Intelligence is a catch-phrase for computer systems that are able to complete tasks that typically in the past have required human intellect. Through natural language processing (NLP), natural language understanding (NLU), and natural language generation (NLG), AI systems are able to read and write and respond to verbal (both oral and written) commands and communications. Through training on incredibly large data sets and deep learning (what are called Large Language Models, LLMs), AI writing systems continue to advance rapidly. AI can generate text in a variety of genres and styles. For this first project, we’re going to focus on the research article summary. With AI, you can take copy-and-pasted or uploaded files or web pages links and ask AI to summarize the text. In some systems you can customize for word length, style, and keyword focus. You can also ask AI to rewrite your own summaries that you have drafted.
Today in class, first with my showing some samples to the whole class and then with you working in a team of three, we will examine and co-write with AI writing systems. [Note: At the time ChatGPT was not as widely available, so I showed students a variety of systems. I probably will still do that in the future, but will focus most fully on ChatGPT.]
AI writing systems freely available in January 2023:
Pre-Activity Check: As noted on the assignment to find an article: Please make sure you can copy and paste text from your PDF. If you cannot highlight text and copy it into a Google or Word file, then it will not work for machine readers. In order to use these systems, the PDF file of your research article must have its text recognized so you can either cut-n-paste from it or you can upload it and the machine can “read” the text.
Article Paste/Summarize: Please run your article through the summarizers and save the outputs by downloading them or copying them. Then please be sure to post your outputs to our class Google Research folder YourNameNAME OF AI SYSTEM1, 2 etc. if you try the same system
twice. Be sure to try several because the writing summary will differ and too it’s helpful to try the same one twice to see if its writing changes each time.
Note Paste/Summarize: In AI writing generators, please run your Notes through saying something like Write a blog post about this information
Paraphrase/Rephrase/ReWrite: Now take the rough draft you wrote and paste that in and ask a system to paraphrase it or to rewrite it for a type of audience. You might also want to try playing with paraphrasers (such as Quillbot Paraphraser quillbot.com)
Copy Editing: You can also enter text to ChatGPT and ask it to correct all grammar/punctuation/spelling and explain to you the corrections it makes.
For homework, you will complete Reflective Memo on AI Writing Systems
6. Reflective Memo on Your Writing & AI Writing Systems
Please write a memo to me (at least one-page, preferably more) where you analyze your writing and the writing produced by the AI systems, reflecting on what AI seemed to do well, what it didn't do so well, where it helped you consider new ideas or new approaches or even just new possible words to use in your writing and too where it was or where you could see it being potentially problematic. Please quote from your draft and from AI's drafts. Please upload your memo here.
Please include as appendices screenshots or cut-and-pastes of some of the outputs from various AI systems, being sure to identify clearly from which system the output is from.
7. Final Draft
Please upload to [our university course management system] one file that will include 3 things:
Evaluation Criteria
As I read and evaluate your final materials I will be considering:
Acknowledgments
I wish to thank the students in English 310 who have dived into writing with AI with such enthusiasm and interest. I also wish to thank Alan Knowles, who while he was a doctoral student and TA at Miami University, led the way on integrating GPT-3 into the course Professional Communication & Digital Rhetoric, generously sharing his assignments with all instructors. And, finally, I wish to thank James Porter with whom I have been researching AI for several years, including in a chapter “AI Writers as Professional Communicators” in our book Professional Communication and Networked Interaction (2017); in a post in the Sweetland DRC blog “The Impact of AI on Writing and Writing Instruction” (2018), a paper “Ethics for AI Writing: The Importance of Rhetorical Context” for AIES '20: Proceedings of the AAAI/ACM Conference on AI, Ethics, and Society (2020); and a paper “Team Roles and Rhetorical Intelligence in Human-Machine Writing” in the Proceedings from the 2022 IEEE International Professional Communication Conference (2022).