Experiment, Reflect, Create: Teaching AI for Public Good

Graduate course challenges students to explore how AI tools can help advance public health

By Frederique Laubepin

Frederique Laubepin is a clinical instructor in instructional design and assistant director of instructional services for the University of Michigan School of Public Health.

What happens when students are given full freedom to use generative AI—but within structures that guide their choices and reflections? In the School of Public Health graduate course PUBHLTH 555, “ChatGPT/AI and Public Health,” students moved from consumers of AI outputs to creators using AI to advance public health. Along the way, they built the literacy, fluency, and meta-skills they’ll need to engage responsibly and navigate a shifting technological landscape.

Building AI Literacy and Fluency

When I was asked in 2023 to design a course on AI and public health, the challenge was clear: how do you prepare students for the future of work in a world already being reshaped by rapid technological change? For me, that meant helping them build critical literacies—things like ethics, responsible use, and a clear sense of AI’s strengths and limits—while also developing fluency with the technology so they could start imagining and creating applications that support public health’s mission. And underneath it all, there was an even bigger goal: helping students build the resilience and the meta-skills they’ll need no matter how the tools evolve—knowing how to learn, staying agile, getting comfortable with uncertainty, trying things out quickly, bouncing back from setbacks, and approaching the whole process with an innovator’s mindset. Easy, right?

Unlike many instructors, I wasn’t wrestling with how to control, constrain, or prevent the use of AI in my classroom. Quite the opposite: AI had to be center stage. The course only made sense if students could work directly with the tools, try them out, and see for themselves what they could (and couldn’t) do.

Of course, that didn’t mean I could just set them loose without any guardrails. Students needed enough structure to stay oriented, but not so much that it closed off exploration, agility, and innovation. They also needed opportunities to step back and reflect—not only on what the tools were producing, but on how they themselves were learning. Those reflections were critical for building the metacognitive skills that would help them adapt, question, and grow long after the semester ended. And applied projects offered a way to connect their experimentation with real public health challenges, grounding the use of AI in meaningful contexts where equity, communication, and impact truly matter.

The Pedagogical Backbone

If some of this sounds familiar, that’s because it echoes established pedagogies—alongside one new, emerging framework. The course wasn’t built from scratch so much as woven from multiple traditions that together balanced freedom with structure.

• Project-based learning (PBL): Students tackled open-ended, applied challenges that gave purpose and meaning to their experimentation with AI. Instead of practicing skills in isolation, they applied them to authentic public health contexts.
• Self-regulated and metacognitive learning: Students set goals, made choices about their approaches, tracked their progress, and reflected on both successes and failures. This turned AI use from a black box into a visible process they could analyze and improve.
• Experiential and inquiry-based learning: Knowledge wasn’t delivered so much as discovered. Students learned by trying things out, asking questions, and testing the limits of the tools—then looping back to reflect and iterate.
• Critical pedagogy: Students engaged with AI not only as a technical tool but as a social and ethical phenomenon. They examined bias, misinformation, and equity issues, and considered how technology use should align with public health’s mission and values.
• Coach for the Approach (Yee, Uttich, Giltner, & Bojanowski, 2025): Though published later, this framework captures the stance I was already taking. My role wasn’t to gatekeep content but to act as a coach, helping students focus on strategies, mindsets, and processes that would carry them forward.

Together, these frameworks formed the pedagogical backbone of the course. They gave students real autonomy and choice while ensuring that exploration was always anchored by structures that built agency, critical reflection, and resilience.

Putting the Backbone Into Practice

Translating this mix of pedagogies into a working course meant designing assignments and structures that gave students both freedom and focus. PUBHLTH 555 was built around four core moves:

Gameful structure with choice and autonomy

Instead of a one-size-fits-all path, students earned experience points (XPs) through a menu of assignments. A few activities were required to build shared competencies, but beyond that, students had autonomy to choose the assignments that fit their interests, strengths, and schedules. This gameful design turned the course into a space where students could explore without fear of failure and chart their own trajectory toward success.

Optional activities ranged from suggesting content for class discussion, annotating readings, and submitting questions for guest speakers to testing and evaluating various AI tools, and digging deeper into topics of interest to each student. In addition, students could earn badges (and extra points) for effort, ideas, and curiosity.

The menu of optional assignments included:

• Pre-class survey (100 pts)
• End of term reflection (500 pts)
• Perusall readings (50 pts each)
• Class-based activities (100 pts per class)
• Content suggestions (50 pts per week)
• Ask the expert (200 pts per guest speaker)
• AI in the news (50 pts per submission)
• Hallucination detective (150 pts per submission)
• Try, test, and evaluate (500 pts per submission)
• AI visualize (200 pts per submission)
• Podcast perspectives (300 pts per submission)

The AI for Health Equity challenge

The centerpiece of the course was a scaffolded, semester-long project, the AI for Health Equity Challenge, where students developed innovative AI-driven solutions to address pressing health equity issues. They began by proposing individual ideas, then peer-reviewed one another’s work before forming teams around shared interests. Teams built prototypes using GPT, Maizey, or CalStudio, submitted technical spec sheets documenting design and ethics decisions, and presented their work in a case competition judged by faculty and alumni.

Reflection and metacognition woven throughout

Throughout the course, assignments encouraged students to track their own processes, recognize patterns in AI’s performance, and reflect on how they were learning. Documenting prompts, comparing outputs, and naming misfires turned messy experimentation into visible learning data—and helped students build the self-awareness and agility they’ll need in a fast-changing technological environment.

Letting go

Perhaps the most important move for me was simply stepping back and letting go of the impulse to tightly control the learning process. I had to trust the students and trust the process: that if I provided the right conditions, they would reach the learning goals, even if the paths they took looked different from what I might have scripted. That meant stepping back from being the sole driver of the course and allowing students’ choices, explorations, and even missteps to shape the direction of their learning.

Student Experiences and Outcomes

The data tell a story of deep engagement. Over the course of the term, each student completed an average of 47 optional activities. Collectively, they earned 81 badges. And almost all of them significantly exceeded the number of points required to pass. This level of voluntary participation reflected not only motivation but also the freedom students felt to pursue what interested them most.

Their work in the AI for Health Equity Challenge showcased both creativity and responsibility. By semester’s end, students had produced 10 functioning chatbots on topics ranging from mental health support and personalized nutrition to elder care, cancer medication management, and water quality education. The winning project, WanderWell, was designed to provide support, education, and harm-reduction strategies for people living with substance-use disorders and their families. Their work was later highlighted in the School of Public Health’s magazine Findings, giving the project visibility beyond the classroom.

At the conclusion of the course, students reflected on what they had learned, how they engaged with the material, and how the experience shaped their understanding of AI and public health. Their comments highlighted the value of student-driven learning, the challenges and rewards of working with emerging technologies, and the impact of connecting coursework to real-world problems.

“The fact that our discussions evolved based on what was happening in the AI world made this class feel so much more connected to real life.”

“I loved that we weren’t just learning about AI in theory—we actually built something. My team’s AI prototype felt like a real contribution to solving a public health issue.”

Together, these outcomes point to a transformation: students not only gained critical literacy about AI but also moved toward fluency and innovation, applying their skills to public health challenges in ways that mattered to them and their communities.

If there’s one thing this course taught me, it’s that students are ready for this challenge. Given the chance to experiment, reflect, and create, they showed not only curiosity but also responsibility and imagination in how they engaged with AI. They leaned into uncertainty, collaborated to solve problems, and connected their work to public health values in ways that surprised even me.

We can’t fully predict how generative AI will evolve, or how it will reshape our fields. But what PUBHLTH 555 demonstrated is that students already have the capacity to navigate that uncertainty—when we design courses that trust them, support them, and give them space to learn by doing.

Practical Tips

Designing a course around emerging technologies doesn’t mean having all the answers in advance. What mattered most in PUBHLTH 555 was creating conditions where students could explore, reflect, and connect their learning to authentic challenges. A few lessons stand out:

• Teach prompting, evaluation, and ethics explicitly. Go beyond showing students how to use AI—help them practice how to think with it, critique outputs, and grapple with ethical questions.
• Balance freedom with structure. Offer scaffolds like checkpoints, reflection prompts, and low-stakes exercises that give direction without stifling exploration.
• Keep it authentic. Frame AI use around disciplinary goals and social responsibility, and ground learning in applied projects that connect to real-world challenges.
• Encourage iteration and inquiry. Design low-stakes, creative activities where students can test, compare, and refine AI outputs, treating AI as a tool for inquiry and imagination—not just efficiency.
• Require reflection. Ask students to document, critique, and analyze how they used AI. Reflection turns tool use into visible learning and metacognitive growth.
• Aim for meta-skills. Beyond technical know-how, help students practice agility, resilience, and comfort with uncertainty—skills they’ll need to adapt as technologies evolve.
• Coach the approach. Step back from tightly controlling the process. Guide strategies and mindsets, help students build resilience and agility, and trust them to reach the learning goals in their own ways.

References

• Deci, E. L., & Ryan, R. M. (2012). Self-determination theory. Handbook of theories of social psychology, 1(20), 416-436.
• Freire, P. (1970). Pedagogy of the oppressed. New York: Continuum.
• Kolb, D. A. (1984). Experiential learning: Experience as the source of learning and development. Englewood Cliffs, NJ: Prentice Hall.
• Markham, T. (2011). Project-Based Learning. Teacher Librarian, 39(2), 38-42.
• Yee, K., Uttich, L., Giltner, L., & Bojanowski, A. (2025). Coach for the approach: The educator’s new role in the age of AI. UCF Created OER Works.

Feature photo illustration by Tim Sharp / School of Public Health