Chapter 7
My Adventures in Teaching AI Fluency to College Students: A Case for Design Literacy
When AI Writes Code, Not Text Outputs
AI literacy lessons mainly concern themselves33 with how to analyze the text output from a chatbot, because it’s a form of information that needs evaluation and citation. That’s information literacy and usually what concerns research librarians about GenAI.
But what about when the chatbot generates computer code? How should we teach our students to think about programming with chatbots? Do students still need to analyze this output with the same critical thinking lens that they do other information outputs?
The main reason why I raise these questions is that this semester, I’ve been an embedded librarian in the first “AI Fluency” class that my midsize university has offered, collaborating with faculty on teaching AI literacy, giving feedback on assignments, and public speaking about the course. At the start of developing the course, the art professors and I believed that teaching this subject meant curating a range of complex, wicked problems that students could use AI tools and services to tackle.
But we soon realized that this approach may not work as we had hoped. To keep students engaged throughout the semester, we instead allowed them to pick a project that let them expand on their passions and interests.
As a result, we’ve had several students create a wide range of websites and apps. We’ve had an indie artist create a website, eye-popping music videos, and AI-generated album artwork. We’ve also had students create an interactive escape room game, a fandom role-playing app, and an e-commerce storefront.
In conversation with students, I noticed that the number one thing that comes up in class is often design process recommendations—what’s more accurately described as design literacy, or the ability to identify a real problem, understand users’ needs, and iteratively prototype solutions—rather than AI fluency.
Do the professors and I provide solid advice on how to write chatbot prompts? Yes. Do we help the students contextualize the latest news and how it relates to their project? Of course. But often, we discuss the latest stage of their prototype and what needs to be done next to create the full-fledged version of their website or app.
We consider the feasibility and usability of their coded project—and how much they can accomplish by next week’s end—because no one really knows the limits of what platforms like Lovable34 and Base4435 can create. So we’re seeing each student do things, create things, and solve problems we couldn’t have imagined when we first came up with the class outline. The students are using a ton of AI, but it’s not to create information.
It’s to create code that results in the design of a website or app.
Thankfully, my degree program at the University of Michigan included classes in user experience design, and I also worked for several years in the tech sector before I pivoted to become a librarian.
I conducted user experience research for a handful of companies, helping them grasp if they developed the right product—ensuring that it was useful, usable, and “well-designed.” So, I have extensive knowledge and professional experience informing my belief that AI fluency instruction is missing a critical foundation: design literacy.
Real-Time Design, Real Blind Spots
The biggest challenge in trying to teach AI fluency fundamentals is that you could easily make a course where students read themselves to death about information literacy and critical thinking. You could demonize chatbots to the point where students fear using them, as I’ve already observed in my position as a Research and Assessment Librarian, where I conduct regular survey-based studies on student and faculty attitudes and opinions on GenAI.
When my colleagues and I started to work on this course, we also believed that our expertise and experience with GenAI would matter more. However, most of these tools and services are easy enough to use, so pretty much anyone can figure out the basic functionality and start producing plausible products and services.
What I’ve started to realize is that when students dive in and start coding, they’re missing the “messy” first part of the design process, wherein you must figure out who you’re building for and what exactly they value. There’s a great concept coined by Lean Startup author Eric Ries called36 “achieving a failure,” which is when you expertly execute a plan that results in creating something that no one actually wants to use or needs.
As such, I’ve started to describe vibe-coding to others as something more akin to “real-time design,” because you’re able to write programming instructions and see them rendered in front-end mock-ups. The trouble is that real-time design can allow you to skip the hardest part of the process: talking to people and trying to understand their environment and experiences.
When you do real-time design, spitting out code and design at the exact same time, you’re taking two things that used to be separate and combining them. I learned this years ago, and it took a long time to wrap my mind around, but a designer first creates the look and feel of an app inside a tool like Figma. That design is then examined to make sure that users can navigate the product, a process called a usability test.
Once that design gets approved, then a front-end developer writes the code that renders the buttons and layout in a way that is now clickable and interactive in the app.
Earlier this week, I started to discuss this challenge with two different colleagues. It revealed how much AI fluency is really a cross-disciplinary field. The first one said that a student is still exercising critical thinking when they decide what to build; the very act of determining what to make should trigger their skepticism of their own ideas. The business professor pointed out that students were making things without trying to determine or discover any kind of product-market fit.
Put simply, product-market fit is the degree to which an app satisfies a strong market demand, and a target market is established.
In other words, I realized that my students were making the same mistakes that are often prevalent in Silicon Valley. Founders overestimate how much people want a solution to a problem all the time—and oftentimes the “problem” is a personal pain point that isn’t shared by everyone. They often don’t conduct user experience research. They don’t go through the design thinking process of empathizing with users, defining a problem to solve, and ideating based on a broad range of solutions.
Essentially, our students were jumping between prototyping and testing mode, without revisiting the other steps in the process. They became like founders who were somewhat blind to their own biases, because they started with the chatbot rather than chatting.
The Front Lines of AI Fluency
The reason I’m sharing this journey and what I’ve learned so far is that I don’t think anyone—let alone us as librarians—knows exactly what it means to teach AI fluency. There are a lot of different ideas out there, and companies have started releasing courses37 on the subject.
But when you’re the librarian supporting a brand-new course—and you’re embedded with the goal of learning and reporting what’s happening at the front lines of AI literacy education—you must stretch yourself and push the boundaries of what you know. You must figure out the gaps in your own logic. I’m lucky in the sense that I’ve had a multitude of experiences and a multidisciplinary education, but I realize that not everyone understands the importance of design literacy.
In a recent column38 in Inside Higher Education, Professor Joseph Squier argues that design literacy is one of the fundamental literacies missing in higher education. He contends that scientific and humanistic literacy currently serve as the primary pillars of undergraduate education.
However, design literacy—what Squier describes as “creating solutions through prototyping, iteration, and synthesis; learning how to transform understanding into action and ideas into reality”—remains crucially and fundamentally absent from most existing education programs. He says that design literacy is an important competency for students to learn and master, as “it’s the discipline of transforming ideas into reality, of imagining what doesn’t exist and bringing it into being.” That’s exactly right.
Design literacy is essential.
Given that research libraries are exploring the creation of AI studios—or what I’ve called “imagination labs” in a previous essay39 —it makes sense that if we’re going to empower students with top-tier models and vibe-coding tools, we also need to help them learn design literacy.
Students must grasp the complete process of how to envision what doesn’t exist and bring their ideas to life, so that they can connect whatever they imagine to the real world and real people who badly need solutions to wicked problems.
Even NN/Group,40 one of the leading user experience education and training firms, has been grappling with how AI coding has changed design work. In a recent blog post,41 Sarah Gibbons and Huei-Hsin Wang talk about how vibe-coding tools allow those with personal intuition and agency to effectively throw the design process out the window, similar to what I observed in the speed at which students prototyped their ideas.
With one exception, no one slowed down and first asked themselves—or anyone else—if any people actually needed their particular solution to their problem.
The Substack Technically recently wrote42 that to understand vibe-coding and all the hype around it, you must look back no further than The Maker Movement and the era of 3D printers, as the parallels are “hard to miss. Technically explains, “Vibe coding has slop. The Maker Movement had crapjects,43 a term the community coined for 3D printed objects that served no purpose beyond proving you could extrude plastic into a shape.”
During my internship in the University of Michigan’s Shapiro Design Lab44 during graduate school, I saw firsthand what kinds of things students created with the 3D printers the library provided. Many printed things were nerdy objects that made people happy; only a small number of objects that patrons peeled off the printer bed were useful and usable, or what I’d describe as well-designed with someone in mind.
To avoid that fate with AI studios—essentially becoming AI slop factories—it’s essential for embedded librarians and professors to design literacy fundamentals. Research librarians can become “trusted guides4 ” who can help students connect what exists in their imagination with problems in the real world. We can help students critically examine their code the same way as a chatbot’s text outputs.
Meanwhile, I believe there’s a tremendous opportunity for librarians to learn how to conduct research, design, and code solutions to their own hard problems. We must become more antifragile: not only withstanding shocks, volatility, and stress, but actually benefiting and growing stronger from the huge and new challenges that navigating the AI age poses.
Here’s my favorite question to ask: What would you build if you deeply understood why people love and enjoy your library, and the role that it plays in their everyday lives?
Build, design, and develop your ideas for them. That’s the powerful and real capacity that’s being unlocked by this new wave of AI-coding platforms.