Univeristy of Pennsylvania
Close

Student as Sorcerer: Using LLMs in the Humanities & Nursing Classroom

Students typing on laptops. Close up of hands, unrecognizable people.

09Jul’24

AI Stories, Feature Stories, Spotlights

July 9, 2024 by Erin Bartnett

"I don’t think Penn students should graduate not having interacted with this thing that will absolutely shape their practice."

During her two years as the national leader of a large nursing organization, Amanda Bettencourt took part in discussions about the future, which often involved discussions about Generative AI’s role in the profession. These conversations inspired her to think about how her students could engage with Generative AI and Large Language Models (LLMs) like ChatGPT before they left Penn: “I don’t think Penn students should graduate not having interacted with this thing that will absolutely shape their practice.” Her goal was to remove fear, but also help students recognize how generative AI can support the clinical work they were already training to do. 

Faculty across Penn are designing assignments that enhance student learning with LLMs, while also empowering students to challenge those same tools using the knowledge they are developing in that specific subject area. Students are encouraged to develop their own thinking, engage critically with the outputs they receive, and challenge or defend those outputs in order to learn more about their field of study. 

Amanda Bettencourt, Simon Richter, and Ann Kuttner have each used generative AI tools as a creative way to empower students to apply what they are already learning. As Simon Richter explained, “The facility Chat has to produce on command…that’s a kind of wizardry. But I want students to be in control of that wizardry. And being in control means understanding what Chat is producing, being in charge of what Chat is producing, and being in a position where you, like the sorcerer, can be in control. Chat isn’t the sorcerer; Chat is the wand.” 

Students typing on laptops. Close up of hands, unrecognizable people.

"Students expressed how valuable it was to read closely and take responsibility for what ChatGPT was generating."

When German ChatGPT is a Lousy Poet  

In Simon Richter’s course, Von handschrift bis Hypertext (From Handwritten Manuscript to Hypertext), students used German ChatGPT to compose and edit creative pieces in short literary genres. Before turning to ChatGPT, however, students empirically studied the literature. Richter provided students with a series of examples for each genre, and they examined and discussed them as a class. This empirical practice was crucial, Richter explained, so students could first “refine [their] sense of what makes up a genre.”   

Then, students began using German ChatGPT. For this assignment, students wrote prompts that asked ChatGPT to compose a piece in a literary genre of their choice; as part of the prompt, students were instructed to include the explicit elements of the genre. Most importantly, they were also told to challenge the initial output they received. Students observed that German ChatGPT’s output felt similar to the way educated Germans wrote, but also more accessible than traditional literary texts. Even for less experienced German students, it was relatively easy to see that ChatGPT wasn’t getting it right. This proclivity for error, Richter explained, inspired students to take the next step—students “are motivated to read it and understand it so that they can figure out what’s wrong.”  

Whether they were working in small groups, pairs, or on their own for homework, students discussed their findings in class. As part of this discussion, Richter asked students to share their experience using German ChatGPT for language learning. While the ChatGPT transcripts students shared were at turns funny and insightful, students expressed how valuable it was to read closely and take responsibility for what ChatGPT was generating. Richter explained, “it motivated a kind of attentive reading that they might, in other circumstances, not practice.”  

"Students needed to develop enough knowledge of the subject to be critical of their findings, whether those findings were an output from ChatGPT or the results of a Google search."

LLMs vs. The Art History Student

How students can use Generative AI to support their learning is an open, ongoing conversation in Ann Kuttner’s courses. In the fall 2023 semester, after doing a deep dive into the available LLMs (ChatGPT, Bing, Claude, etc.), Kuttner asked her Greek Art and Artifact students how they had been using generative AI and encouraged them to share their experiences.  

In these conversations, she introduced students to LLMs they might use to jumpstart their learning. Kuttner offered a list of the free LLMs and provided some context on which would be most useful for specific assignments in the course. For example, she suggested they try using ChatPDF to get started on more complicated reading assignments or ask another LLM about the different theoretical approaches they might take on a given subject. However, she also emphasized that, as with any other research tool, using LLMs isn’t a replacement for thinking and learning, nor was it required that students use LLMs in her course. Students needed to develop enough knowledge of the subject to be critical of their findings, whether those findings were an output from ChatGPT or the results of a Google search. 

In order to help students see the value of their own observation skills, she created a new description assignment, inspired by her colleague, Huey Copeland. Students wrote a paper (without using LLMs) that included at least one strong analytic description of an object from the course. Then, students could ask an LLM of their choice to produce a description of the same object. Students submitted the LLM output and their own reflection on the differences between their description and the LLM’s. Generally, Kuttner said, the students’ findings were encouraging: “[Students] were very shocked that it often didn’t know their [object]. Often, they said that their [description] was actually pretty good, or got to things the bot didn’t get to, which is a cheering and empowering moment for them.”  

"I want them to think about what a human caregiver can provide to a patient or family member who has a question or health concern that a computer algorithm can’t."

Using Generative AI to Prepare Future Nursing Leaders  

Amanda Bettencourt continues to find ways to empower her nursing students through assignments that leverage LLMs. Over the course of a few semesters, she developed an assignment that asked students in her pediatric clinical course to use generative AI to answer a question they had heard from a patient or family member from their clinical experience. She wanted students to evaluate “what the AI knows and what it doesn’t know, and what [the students] know as nurses that the AI doesn’t know.” 

While part of the assignment goal was to get students more comfortable using Generative AI tools, Bettencourt explained that she really wanted her students to think about their future role as nurses: “I want them to think about what a human caregiver can provide to a patient or family member who has a question or health concern that a computer algorithm can’t.”  

This is the third year she has used this assignment in her course. For some students, it was one of their favorite parts of the class. Bettencourt recalled the experience of one student interested in exploring how ChatGPT would approach a disease process they had seen in clinical. While ChatGPT was able to provide treatment and medications, it failed to list any holistic care to support the child’s development and growth. She said it was the moment when the student saw “all the pieces of nursing care that are part of our profession but not ‘Googleable.’” For that student, using ChatGPT put “an exclamation point” on a lesson Bettencourt had been teaching all semester long: “how nurses promote the health of children in society.” 

If you are interested in learning more about whether or how to use Generative AI tools in your own teaching, visit CETLI’s web resource, Generative AI & Your Teaching, sign up for the Seminar, Using AI Productively for Teaching, or email us to brainstorm about specific questions you have for your own courses.  

To learn about how faculty in STEM fields are using Generative AI tools like ChatGPT to help students in introductory courses learn advanced topics, read the Feature Story, Asking the Right Questions. 

News & Features

More Resources