GPT Field Notes
GPT Field Notes
Share an AI teaching tip, help create a resource!
With the advent and rapid, ongoing development of artificial intelligence tools, we are inviting UNC instructors to share an example of a teaching practice related to AI technology that’s working in their courses. Write a short paragraph to describe an instructional practice, course policy, or other teaching-related use of AI that you think has enhanced student learning or contributed to your success as an instructor. Browse the examples below by category.
Share your paragraph through our simple submission form. We will continue to update categories as we receive submissions.
Students (working in groups of four)were given a short easy paper assignment and asked to use two different AIs and two different prompts for each AI to write the paper. This was to be followed by a critique of each AI response and a paper of their own. After this was handed in, we had a class discussion about the strengths and weaknesses of the AI responses. This helped the students be more aware of the need to evaluate AI responses and how and when AI is useful. I allow students to use AI in papers but they must state that they used AI and give the prompt and the AI response at the end of thier paper after the citations.
Ann Matthysse, ProfessorBiology
BIOL481, Biology, History, and Evolution of Epidemic Disease
Enrollment: 20
In my intro-level journalism, the use of uncredited and/or unauthorized AI is prohibited. Yet many UNC students were cheating. After explaining to the class that some community members were turning in assignments that appeared to be entirely AI-generated, I gave everyone an in-class mandatory scored-but-ungraded quiz to check their knowledge regarding the MEJO 153 syllabus rules along with UNC rules from the Office of the Provost, which outlines expectations for responsible AI use and what constitutes academic dishonesty. I wasn't sure if students were cheating because they didn't understand rules, but they did: most scores were above 95. Afterwards, we talked through how other classes and professors permitted AI use, which was chaotic and all over the map. Then we voted on retroactive consequences for AI-related academic dishonesty in a Google poll with 10 options. The majority vote won. Cheating subsided.
Erin Siegal McIntyre, Assistant ProfessorJournalism
MEJO 153, News Writing and Reporting
Enrollment: 18
In MEJO 652, Digital Media Economics, the main project for the semester is to study a media company from its finances to its leadership and conduct a competitive analysis. As part of the project, students study the use of artificial intelligence in media organizations from across the world. We discuss the use of AI news anchors to AI-produced sports stories to AI use in investigative data reports. We discuss ethics and transparency and current laws related to content creation and copyright. Additionally, we focus on the of AI for customer service, data analytics, social media marketing and other business-related functions. As part of their project, students must assess the use of AI in their organization currently and recommend future uses as part of their overall media assessment. All of this helps them to better understand and assess media uses AI to increase audience and revenue.
Marisa Porto, ProfessorHussman School of Journalism and Media
MEJO652, Media Economics
Enrollment: 42
I demonstrate how to use Chat GTP to find evidence about a specific topic and show how hallucinations can happen! I also show how to use ChatGPT to brainstorm research topics/ PICOT questions in nursing and innovations in health care. If students do use ChatGPT for their paper, I ask they fill out a transparency table on how they used it and provide links so I can see prompts and the final product. We do address the power of critical thinking and how AI can rob one of that skill- and how to respond to patients/families who come with AI-generated information. Students often choose to research AI-infused innovations in their health innovation epubs and make an evidence based recommendation about its current use. We also use a great AI-enhanced app called YOODLI to give communication feedback to undergraduate and graduate nursing students.
Maureen J. Baker, Associate ProfessorSchool of Nursing
N302, Research, Ethics and Innovation
Enrollment: 114
In SOCI57H, my first year honors’ seminar, many of my assignments incorporate ChatGPT use. For their term paper, students must pick an occupation or profession and explain how it has been rationalized (or not) over the past century. They use Microsoft’s Copilot/Bing as the main search engine because it is free and connects to the Internet, as well as make use of its generative AI abilities. For example, one assignment asks them to prompt Bing to give them evidence-based descriptions of what work in that occupation would be like 50 years ago and today. They follow up Bing’s initial answer by asking about how work would differ by a worker’s gender and race. Rather than set restrictions on the use of generative AI, I am encouraging it and helping students to engineer better prompts for the Chatbot.
Howard Aldrich, Kenan Professor of SociologySociology
SOCI57H, The Rationalization of Social Life in the 21st Century
Enrollment: 24
We are doing a 10-week project to investigate the microbiome using Generative AI. Students are required to fill out a form indicating which Chatbot they used (they are recommended to start with Bard, Bing, or ChatGPT). They submit the two prompts they have used to get the information for that week. They are also required to get three citations that back up the information generated. They must link the citations into the form, and prove that they are accurate. If the citations are inaccurate (hallucinations), they must find an appropriate citation that either supports or disproves the material generated. They then discuss not only their (way cool) information regarding the microbiome, but how effect the AI that they uses was, and how accurate the citations. Then they grade one other student's submission for accuracy and citations.
Lorraine Cramer, Teaching Assistant ProfessorMicrobiology
MCRO 251, Medical Microbiology
Enrollment: 226
On the eve of the last day of class, I asked Chat to prove about 12 mathematical statements, all of them pulled from homework assignment or our review sheet. I intentionally included one false statement that had been assigned in error to the students. I accidentally included one additional false statement when I typed in the wrong inequality sign, and then I also asked ChapGPT to prove the correct version. On the last day of class I brought in a transcript of the conversation that included 8 of the "proofs" that had errors in them. I asked students to pick two to read, and to comment on what ChaptGPT did well, what it did poorly, and to pinpoint exactly where the logical errors were in each proof.
Linda Green, Teaching Associate ProfessorMathematics
MATH 381, Discrete Math
Enrollment: 35
A Teaching Assistant provided an overview of Chat GPT following a lecture about how to complete patient documentation. Students have assignments following this lecture in which they write a patient note based on case details provided. Students have been given the option to submit one assignment using Chat GPT to generate the note. If they select this option they must include their prompts used to generate the note and a critique of the quality of the note generate.
Lisa Johnston, ProfessorSOM – Health Sciences
PHYT 720, Introduction to PT Exam and Intervention
Enrollment: 32
One work product for this class is a10-page research paper on the topic of an ethical controversy in the field of genetics. Many of my students (mostly senior Biology majors) have not written an essay since high school, so we devote several class sessions to scaffolding the process. This year is the first time I've had to confront the reality that the learning objectives for this project do not align with use of generative AI. After struggling to decide how to pivot, I decided to put it in the hands of the students. I implemented a multi-part consensus-building process, which included prework reading, in-class discussion, anonymous forum posts, and finally synthesis of a coherent policy. My hope is that having it be student generated will make it more relevant (and thus more likely to be followed). The process itself was also a valuable learning experience for students about how to navigate uncertainty.
Lillian Zwemer, Teaching Assistant ProfessorBiology
BIOL 405, Good Genes: Human Reproduction in the Social Context
Enrollment: 35
I've been using generative AI in one of my pharmacotherapy courses to create treatment plans in response to patient cases. Those treatment plans -- which often contain inaccurate, misleading, or incomplete information and frequently fail to account for nuanced patient-specific information within the case -- are then provided to my pharmacy students. The students are instructed to critique the treatment plan, recommend changes, and provide supporting information to justify their recommendations. The activity also serves as an opportunity for the students to reflect on the potential benefits and limitations of using generative AI as a tool within healthcare settings.
Zachary Noel, Teaching Associate ProfessorSchool of Pharmacy
Multiple course numbers, titles
Enrollment: Various
November 7th: DUE: Chat-GPT Assignment (~500-1000 words)
Write a basic prompt for GPT-4 (https://openai.com/product/gpt-4) using the formulation
“Analyze _____ in 500 words.” (For example: Analyze contemporary telephone use in 500 words.) You are encouraged to write a (more) specific prompt about a topic that interests you.
Copy the response into a word document and use “track changes” to provide comments and/or write a separate commentary on the response.
In your comments or commentary, note the claims in the response that would, in a scholarly research paper, require sourcing or documentation.
Peer review workshop of papers in class:
Bring a paper copy or a phone/tablet/laptop to access a classmate’s paper.
Michael Palm, Associate ProfessorDept. of Communication, CAS
COMM 453, The History of New Media Technology in Everyday Life
Enrollment: 30
In MEJO 187, students are required to use AI to generate content for website mockups (images, text), and they're encouraged to use it when writing code. However, they must disclose their AI usage, and they must include the links to their AI chatbot conversations as part of the assignments. From reviewing the chatbot threads, I can see how students are writing prompts, I can see how their prompts have changed as they've gained more knowledge on the topics, and I can offer advice on using AI effectively. It's been fascinating. It's helped me incorporate AI into class exercises, e.g., I can show examples of effective prompts and ones that don't generate good responses. It's also revealed new ethical gray areas in content creation (e.g., is it okay to use AI to generate portraits of real people who aren't recognizable public figures? Or fake people that AI deems "representative" of a culture?).
Scott Geier, Teaching Assistant ProfessorHussman School of Journalism and Media
MEJO 187, Foundations of Interactive Media
Enrollment: 7
We incorporated one of the course assignments on providing patient health education at the appropriate health literacy level using AI software. We had the students use a specific instructor-provided prompt and then had the student compare the results with a scholarly patient health education source.
Lena Brokob, Clinical InstructorSchool of Nursing: Undergraduates
NURS 512, Care of Patients and their Families with Cancer
Enrollment: 40
I've used ChatGPT for both code writing and essay/grant writing exercises in class. On take-home exams I allow the use of GPT and ask students to write reflections of their use. I've also allowed students to use it to write mock feasibility proposals with appropriate citation of their use of the technology.
Stefan Jeglinski, Teaching Associate ProfessorPhysics and Astronomy
PHYS231, Physical Computing
Enrollment: 25-30
This summer, I am including a course policy on the use of ChatGPT in my syllabus, nested under the Honor Code statement. It reads: "If you consult ChatGPT in the process of responding to exam questions or completing homework assignments, you must cite it as a source, like this: (ChatGPT, personal communication, May 31, 2023). Failure to do so is a form of plagiarism and I will consider it a violation of the honor code. Along with a citation, I would like you to include the prompts that you gave to the system and a brief (1-2 line) explanation of how you used the information it produced."
Emily Boehm, Visiting Assistant Teaching ProfessorBiology
Biol278, Animal Behavior
Enrollment: 50