Duke educators and researchers built Pressing Prompts, an open resource meant to help instructors and students wrestle with the hard questions surrounding AI. They used AI itself to build it.
Three people, who don’t code, set out to create a digital resource that would help students and instructors think critically about AI. They built it using AI. That’s the short story behind Pressing Prompts, an open-access collection of activities, resources, and teaching guidance for college instructors who want to bring AI ethics into their classrooms.
A little background
The project didn’t start out as Pressing Prompts. It began at Duke University during the 2024–2025 academic year as a collaboration between Duke’s Center for Teaching and Learning and Duke Libraries, under the name the AI Ethics Learning Toolkit, co-designed by undergraduates, librarians, faculty, and researchers.
Co-founder Aria Chernik, Associate Professor of the Practice at Duke’s Social Science Research Institute and Assistant Vice Provost for Academic Innovation at the Center for Teaching and Learning, says the whole thing started as a Google Doc, then a board game idea, then a project handed to a web developer who built it into the original AI Ethics Learning Toolkit. Two undergraduate research assistants helped curate the first set of topics.
Chernik was the one who spotted the gap in the first place. “There’s all this need at Duke, and faculty keep asking about ethics,” she explains. “We should create a resource that’s really interesting and beautiful and different.” She’d already been co-teaching with Hannah Rozear for years, so she knew who to call first. “Hannah’s amazing,” she says. “Let’s bring her in.” From there, the group grew into the full team: Chernik, Rozear (librarian at Duke Libraries who now serves as the AI learning librarian), and Remi Kalir (researcher and Associate Director at the Center for Teaching and Learning), who shaped the original toolkit together. Eventually they wanted more flexibility than that first version allowed, so they rebuilt it from scratch, and Pressing Prompts became an openly accessible teaching resource.
What does it offer instructors?
Say a professor is curious about a question like “Is AI biased?” Pressing Prompts opens with an overview of the topic, then lays out learning objectives to build from. From there it offers activities, some as simple as conversation starters, others deeper dives into the subject.
There are also discipline-specific extensions. Click into “Is AI biased?” and scroll down, and you’ll find versions tailored to different fields, including linguistics and art and design, so instructors can connect the big questions to whatever their students are studying. A playlist feature lets them string several activities together into a single lesson.
The project is rooted in critical pedagogy: curiosity, dialogue, reflection, shared humanity. The team is clear that they don’t want education to simply train students to use AI without question. Instead, they want students to examine how power, labor, authorship, and trust show up in everyday technology use, AI included.
It’s also deliberately not a rigid instruction manual. It’s built for dialogic, experiential learning, meaning the hope is that instructors and students keep experimenting and testing the limits of what’s pedagogically possible, rather than following a script.
Privacy mattered to the team too. There are no accounts, no sign-ins, and no personal data collected from visitors.
A librarian’s read of it all
Hannah Rozear, Duke’s Librarian for AI Learning, came to the project through what she calls “a lucky coincidence.” Her role put her in a position to team up with the Center for Teaching and Learning on an AI literacy project. “We liked the idea of presenting educators, and students, with a really wide range of AI topics that touched on big, almost existential questions,” she says, “like, can we trust AI? Is AI theft?” The team settled on building “adaptable, interdisciplinary materials for instructors to explore for their own class contexts.”
Rozear’s background as a librarian gives her a distinct vantage point on what generative AI means for students and educators. “AI-generated information is now everywhere,” she says, “which makes being a diligent fact-checker more important than ever.” It’s easy to forget that AI “is not a search engine, or interactive encyclopedia,” and its bias and inaccuracy can be subtle. She also points to AI sycophancy, the tendency of these systems to want to please, which means students who lean on it heavily “may not be challenged as much about their ideas,” even though that pushback is often where real learning happens. Still, she sees genuine upside. The availability of AI lets students get help at any point in their research, and she’s heard plenty of students say it helped a difficult concept finally click.
What surprised Rozear was watching which topics the undergraduate researchers gravitated toward. “Does AI Harm Critical Thinking?” turned out to be a favorite among the student team, and it’s now the site’s most-viewed page, ahead of even the landing page. As Hannah puts it, “clearly it’s a topic on people’s minds!” She’s also noticed how carefully students weigh their own reliance on AI, “because they recognize that there are many skills they need to learn the hard way, and they don’t want to take shortcuts.” And the audience for the material has turned out to be far broader than she expected: “In the past few months, I’ve presented to the Duke Neurology department, K-12 teachers, and retired Catholic nuns!”
A researcher’s take on why it started, and why it matters
Remi Kalir, the project’s co-founder and Associate Director at the Center for Teaching and Learning, traces the whole effort back to “student curiosity and student concern.” Two undergraduate research assistants, Emma Ren and Barron Brothers, sat on the original design team while running their own research into how Duke students use and perceive AI. The questions driving Pressing Prompts today, including AI’s environmental footprint, its effect on critical thinking, and its biases, are the ones they were already wrestling with. As Kalir puts it plainly: “It started with students.”
Plenty of sites offer AI lesson plans, and a few even focus on AI ethics, but Kalir argues Pressing Prompts stands apart. It’s co-designed by students, staff, and faculty together, it covers subjects like surveillance, labor, and disinformation that are “often absent in other resources,” and it invites inquiry through multiple kinds of activities rather than a single format.
“Very few of us are actually experts in AI and education,” he says. “We grow through shared inquiry and the acknowledgment of our ignorance, not through the prescription of rote answers” to hard questions like “Who benefits from AI?” and “Is AI a friend?” Posing these “pressing prompts” is his way of admitting that everyone should stay curious and open-minded. It’s also, he notes, “an invitation for others to tell us what matters to them.” Colleagues are welcome to suggest questions they think are missing, and the team plans to keep researching and building the project out from there.
Asked to look five years down the road, Kalir says, “Making predictions is an easy way to lose clarity about what matters now.” Bias in AI matters now. Labor exploitation matters now. Environmental sustainability needs addressing now. Pressing Prompts, in his view, exists to help instructors and students “better grasp this moment and make informed decisions about how their learning and their lives are shaped by, and can shape, AI.”
AI helped build the site about AI ethics
None of the three co-founders consider themselves coders. As Chernik explains, “we did this with Claude.” Claude helped with the content architecture, organizing themes and topics, the visual identity, including illustrations, icons, and typography, a full web accessibility audit, and the technical build of the site itself.
The team sees that as part of the lesson, not something separate from it. To wrestle with a question like “Do we need AI?” you have to actually use it yourself and see where it helps and where it doesn’t. That’s why they’re upfront about how the site came together.
Chernik credits the whole project to the people involved, a librarian, a researcher, and an educator, working together. “We could only create this because of the expertise and human knowledge that we have coming to this,” she says, building something “rooted in academic questions.”
Where it’s headed
Pressing Prompts is free and open to anyone, though it was originally built with faculty in mind. Chernik notes that K-12 educators have already started reaching out about it too. This fall, the Duke team is launching a pedagogy series and workshops to formally introduce faculty to the resource, with the hope that instructors will use it in their own classes and contribute back, turning it into a genuinely living, open education resource rather than a fixed one.
There are already signs that’s happening. Chernik recalls one moment that stuck with her: a colleague in occupational therapy mentioned, almost in passing during an unrelated meeting, that she’d taken the earlier toolkit and created a version of it for nursing students.
Chernik and Kalir are also considering teaching a class in the spring where students would contribute directly to Pressing Prompts. It’s the kind of participation they hope to see more of as the resource grows: faculty and students using it in their own settings, adapting it to their needs, and adding to the conversation.
The goal is to create a place where faculty and students can ask better questions about AI, explore what it means for their work, and figure it out together.