LITS Interns Build AI Tools for Teaching, Research

This past summer, two students worked with Bryn Mawr faculty to build specialty AI tools for teaching and research, providing a model for how the College can integrate the technology while remaining true to its mission. The two students, Pragya Silwal '27 and Yumna Fatima Dar '29, worked in LITS as part of the Digital Technology Internship (DTI) program. "Since AI is such a complex and rapidly changing field, Pragya and Yumna were given a real challenge with these projects, which were unlike anything LITS has previously done with professors," said Jeff Hopkins, Educational Technology Specialist in LITS and supervisor for the internship. "However, they more than rose to it. I was so impressed with the work they did throughout the internship and the tools they ultimately built."

For their part, Silwal and Dar found the internship to be a useful experience. “I was very delighted to be a part of LITS’s and EAST’s very first efforts to explore how AI could become a resource on campus,” Dar said. “The internship challenged me in ways that I didn’t expect: me and Pragya spent a lot of time learning new tools and technologies and figuring out how to translate professors’ specific needs into useful tools. But that is also what made the experience so incredibly rewarding.” Silwal added that, “The best part for me was the unbounded room for exploration. Every day, Yumna and I were able to hop on a learning adventure and collect ideas to experiment with. This was not just fun, but also instrumental in shaping the vision of our AI tools and defining what’s possible at an institution like Bryn Mawr.” 

On many projects, Silwal and Dar used BoodleBox, a "compiler" tool that lets users interact with a range of AI services under a single account. One feature lets users develop custom bots, which Silwal and Dar used to build two Retrieval Augmented Generation (RAG) chatbots that could serve as virtual TA's for two courses, one taught by Greg Davis, Associate Professor of Biology, and another taught by Xuemei "May" Cheng, Professor of Physics and the Rachel C. Hale Professor in the Sciences and Mathematics. RAG chatbots, which have custom instructions and access to specialized materials, are designed to answer user questions on specific subjects, rather than be a general-purpose tool. Silwal and Dar used Davis and Cheng's respective teaching materials to create RAG chatbots for their courses. Both professors hope these bots can enhance students' experience by serving as a 24/7 resource for answering questions and assessing learning. “My initial instinct was simply that we needed to try to harness AI to improve learning rather than simply allowing it to wreak havoc,” Davis said. "As I got into it and spoke with colleagues, the challenge came more into focus: how do we create an AI tool that supports the learning process rather than simply offering up short cuts?” To do just that, the DTI's made sure the bot guided students to their own answers by posing Socratic-like questions and providing links back to class materials. "With all these projects, we're trying to build tools that deepen students' engagement with and understanding of course materials," said Hopkins. "Pragya and Yumna did amazing work fine-tuning of these bots to make it so they would enhance, not diminish, students' educational experience." Cheng will introduce her Virtual TA this fall in her course, Physics 121: Modern Physics. Meanwhile, Davis will roll out his tool when he teaches Biology 236: Evolutionary Biology this spring.

Silwal and Dar also used BoodleBox to create a tool for Introductory Physics I: Physics 101, a part of this year's postbac program, taught by Asja Radja, Assistant Professor of Physics. For the course, students complete a pre-assessment of their quantitative skills. From there, they can enter their results into a site built by Dar, which will subsequently provide links to publicly available study resources on the subjects where the students struggled most. The site will also have a link to a “Q bot”, created in BoodleBox, that can further analyze results and provide feedback. “This project will help us identify gaps in students’ prerequisite math skills before those gaps become barriers to learning physics and connect students with targeted resources and personalized feedback,” Radja said. "I could not have developed this tool without Pragya and Yumna, whose creativity, technical skill, and dedication were truly exceptional.”

The interns didn't just create AI tools in BoodleBox, however. They also developed two tools using open-source AI models, which can be downloaded onto computers rather than accessed through the internet. The first tool was built for Visiting Assistant Professor of East Asian Languages and Cultures Heejin Kim, who's creating a Korean open-educational resources (OER) textbook for a Digital Bryn Mawr grant. For the book, Kim wanted a chatbot that could provide students with realistic scenarios in which to practice their language skills. Dar was able to achieve this by combing two large-language models, one in English the other in Korean. "The chatbot Silwal and Dar developed will give students access to the kinds of real-world Korean they rarely see in textbooks,” Kim said. “It's an impressive tool that will meaningfully expand how our students learn and practice the language." Kim will keep working on the book this year, with an eye towards publishing it in the spring. Meanwhile, Silwal worked with Tom Mozdzer, Professor of Biology, to create a visual large language model tool that could turn handwritten field notes into spreadsheets. "The newly developed tool will save us hundreds of hours of manual data entry and allow us to move much more quickly from field observations to data analysis,” Modzer said. “I’m especially excited about its potential to streamline how we process handwritten field notes and make those data immediately usable for research." While their respective projects challenged their abilities, both found the work engaging and rewarding. “I really enjoyed getting to research different tools and figure out how we could use them to come up with creative solutions for each project. It pushed me to explore areas of AI that I hadn’t worked with before and really broadened my perspective on what these technologies could do.” Dar said. “Given how vague the use cases for AI are, the entire summer was a constant cycle of identifying new tools, learning how to use them, and iterating our products. As challenging as it was, it was extremely rewarding and I’m so excited to see these tools roll out on campus,” Silwal added.

While Silwal and Dar have finished the internship, they hope to keep working on some of these projects throughout the year. Beyond that, they think the internship will be a springboard to what they want to do after college. “I’m really glad I had the opportunity to be part of this internship, and I would love to keep contributing to projects that explore how AI can be used in thoughtful and responsible ways.” Dar said.” “This internship has exposed me to how exciting the process of making something happen is, specifically those that positively affect so many of us. I’m grateful for the opportunity and more importantly Jeff’s faith in us to figure stuff out," said Silwal. “I’m always impressed by Bryn Mawr students, but Pragya and Yumna really went above and beyond," Hopkins said, "I can't wait to see how they keep using the skills they've built this summer."