Pillar 4: Student AI Literacy & Storytelling

Whether you work in film, marketing, healthcare, automobile, or real-estate, generative AI is changing the way your job is executed, and those who adapt early will reap its benefits sooner. All professions will be affected by generative AI. Its invention can be compared to the invention of photography, a true creative revolution. If you want to be part of the leaders that are advancing this revolution, this course can get you started on your learning journey.

In this course, generative AI expert Pinar Seyhan Demirdag covers the basics of generative AI, with topics including what it is, how it works, how to create your own content, different types of models, future predictions, and ethical implications.

Check Out This Course

Generative AI is reshaping industries by transforming how work gets done. This learning path puts you at the forefront of this creative revolution, covering content creation, key models, future trends, ethical considerations, and tools like Microsoft Copilot. Gain essential skills to harness generative AI in your career. Start today so you can lead in this transformative era.

Check Out This Course

AI & the Problem of Tacit Knowledge: Or: How (Not) to Ask an LLM a Question: Written by Nathan Szadowski

Intuitively, it would seem that the best way to get a straightforward answer from an LLM is to give it a straightforward prompt, right? For some things, I think that can work—(‘How deep is the Pacific Ocean?’, ‘What’s 5+5?’, ‘Respond to this discussion post, make no mistakes:..’, etc.)—but a lot of the time, it just doesn’t. In my experience, the most egregious issues with current models are unearned confidence, instinctive deference, and a general lack of tacit knowledge. In other words, it misinforms you with the same tone it uses for the truth, it makes judgement calls based on what you already seem to want (even if what you want is dumb), and it only answers the question you asked, not the one you meant to ask.

In the summer of 2025, I had the opportunity to work with an educational technology startup building AI-driven software for academic counselors. On the surface, it was pretty simple: data from PowerSchool would be filtered through a web application, allowing advisors to log meeting notes, review academic trends, and quickly navigate student records with natural language queries like ‘List all sophomores with a GPA over 3.5’. My job, as a developer, was to build the AI part, which I soon discovered would not be pretty simple.

The first instance of the AI assistant was polite, cooperative, and eager to help. Only, there was a slight problem with its responses: they were nearly all wrong. Even though I’d given it the ability to query the PowerSchool data, it hadn’t quite worked out when it was supposed to do so. Thankfully, adding a couple sentences to the system prompt resolved the issue, but the part that concerned me the most was that the data in its responses, despite being completely fabricated, had the appearance of being correct. The only reason I flagged the issue at all was because I had the real data pulled up in another tab to double check. The LLM presented blatantly false data with absolute confidence, giving no indication of uncertainty or confusion.

The second major iteration was somewhat useful; it could perform basic lookups, like ‘What grade is John Doe in?’ and it could even give reliable answers to longer, more specific prompts. Where it fell short was on requests that sounded simple, but required a more complex, implicit series of operations or data to complete. I’d ask it to calculate a student’s GPA, and it would return the weighted value for the current term, even though I meant for it to be cumulative and unweighted. To fix this, I decided to extract the more advanced workflows (like GPA calculation) into separate, deterministic tool functions that could be called by the model. After that was taken care of, the assistant became far more helpful, but it required a lot of extra trial and error to get it to that point.

We also made use of AI as a sounding board during the development process, which revealed a more structural problem: confirmation bias. When asked to make a judgement call, the model would instinctively defer to the opinion of the prompter. For example, I’d come up with an idea for the implementation of a feature, and upon asking the LLM if it was any good, I’d receive almost universally positive feedback. Often, I’d then move forward with that approach, only to find that it either wouldn’t be adequate for our use case, or it wouldn’t mesh well with the existing codebase at all. The other side of this is that, when presented with the exact same idea, but instead framed as something I’m planning to scrap, the LLM would immediately point out all the flaws I’d already discovered during my attempt to implement it (and even a few that I hadn’t yet noticed!) Asking it to be harsh during the initial evaluation didn’t work either, as it produced mostly performed harshness, which is just the same issue in a different font. In the end, it became clear that the model couldn’t act as a check on my own judgement, which is most of what I’d been using it for.

All this is to say, as someone who’s worked pretty extensively with AI, the best way to get value out of its answers is to avoid assuming it understands your questions. Whenever possible, give it finer, more detailed constraints than you think it should need, even if that means writing your own system prompt or, if the situation calls for it, learning how to make your own custom tools. Perhaps more importantly, before you ask an LLM a question, decide how you would find out if the answer were wrong. The common thread here is that all of these failures required external, human-made sources of tacit knowledge to detect. The more familiar you are with those sources—whether the textbook for a difficult course, background knowledge on a particular field of study, or just your own personal insight—the better you’ll be at parsing the AI’s response and filtering out all the nonsense.

AI Beyond Butler: Post-Graduate Experiences: Written by Sarah Miller

The first time I intentionally used a generative AI tool was during my junior year at Butler. I remember feeling hesitant, not necessarily about what AI could do, but about how I could use it in a responsible and authentic way. Just over a year and a half later, AI had become an important part of how I worked, designed, taught, communicated, and problem-solved after graduation. Butler gave me a foundation not only in the technical skills I would eventually use with AI, but also in critical thinking, communication, creativity, and problem-solving. My experiences after graduation gave me the opportunity to practice those skills in environments that were very different from the classroom.

AI in Design

During the summer of 2025, I spent two months as a UI/UX intern with MiAltar, a startup in Lisbon, Portugal. I had just graduated from Butler with a degree in Web Design & Development, and I was excited to gain hands-on experience in the field. My academic background had prepared me with UI/UX principles, design practices, and project management skills. However, I was quickly introduced to an AI tool that changed the way I approached the design and prototyping process: Lovable.dev.

With a prompt, I could quickly explore different layouts and possibilities, and those prototypes became conversation starters with the team. The most valuable part of the experience was learning how AI could bring people with different areas of expertise into the same conversation. I had more technical and UI/UX knowledge, while the co-founders brought their cultural knowledge, passion, and business experience. AI became a shared language for expressing ideas and turning creative visions into something we could all see, discuss, and refine.

AI as a Communication Tool

After finishing my internship in Lisbon, I spent the 2025–2026 academic year in Lyon, France, as an English Teaching Assistant with the U.S. Fulbright Program. With a strong background in French, I felt confident in my language skills and cultural knowledge when I arrived in France…then I had to navigate real life.

One of the biggest surprises of living abroad was realizing how different classroom French could be from the language I encountered in everyday situations. AI translation tools became one of the ways I navigated those moments. AI could help me understand what something meant, identify language I didn’t know, or help me formulate a response. but I still had to bring my own cultural awareness and judgment to the interaction.

This experience also changed the way I think about AI and language learning. I didn’t use it as a replacement for learning French. Instead, it gave me the support I needed to participate more fully in situations where my language skills were still developing.

AI in Teaching

As an English Teaching Assistant, I taught daily lessons to a range of high school students in Lyon. I brainstormed topics, researched ways to make them engaging, and designed activities for the students I worked with. That planning process gave me a strong foundation, but AI allowed me to take it one step further.

Tools such as ChatGPT became useful brainstorming partners. I could provide information about the students’ language levels, the topic I wanted to teach, the amount of time I had, and the type of activity I was hoping to create. AI could then help me brainstorm examples, discussion questions, vocabulary, and activities, but the output was never the finished lesson. The most useful AI-generated ideas were the ones I could evaluate and adapt based on what I actually knew about my students.

Overall Thoughts

It’s easy to talk about AI in terms of what it can produce, but my experiences after Butler have made me much more interested in what AI allows people to do. AI gave me more room to experiment with design ideas, communicate across language and cultural barriers, and adapt lessons for students. However, none of those outcomes came from AI alone. They came from the combination of technology, human expertise, creativity, curiosity, cultural awareness, and critical thinking.

Butler gave me the foundation to think critically and creatively. My experiences in Lisbon and Lyon gave me opportunities to apply those skills in completely new contexts. Now, working on the OEET team, I have the opportunity to help others think about what AI integration and education can look like for the next generation of Butler students.