Recognizing the looming impact of artificial intelligence on the current landscape in higher education and libraries, the University of Texas Libraries has been experimenting with an AI-driven chatbot that could eventually augment library staff to ensure continuous availability of assistance at times when our staff are not available.
The project, which serves as a research initiative rather than a production service, explores the potential of artificial intelligence to enhance user experience.
This exploration has been spearheaded by Hannah Moutran, a recent graduate and Library Specialist, in coordination with Aaron Choate, the Director of Research & Strategy at the University of Texas Libraries.
Choate has been instrumental in organizing an AI interest group aimed at educating library, archives, and museum staff about the potential and applications of AI. Moutran, whose academic research has focused on AI implementations, was a perfect fit for leading the chatbot project, which emerged from discussions among the principals in the interest group.
The chatbot, envisioned as a backup for the “Ask a Librarian” service, is being tested to understand its capacity to provide uninterrupted assistance when human staff are unavailable. To build this system, Moutran analyzed chat logs from the fall 2022 semester, gaining insights into the types of questions users asked and the responses provided by librarians. This analysis revealed that users were often referred to other librarians, departments, or websites for more detailed information.
The chatbot’s development has been guided by a series of in-depth interviews with five librarians from the “Ask a Librarian” service. These interviews uncovered several design priorities for the AI system, including transparency about its nature and data usage, the accuracy of information provided, and alignment with the library’s mission of fostering human connections. Ethical considerations were front and center in the design, ensuring that users know when they are interacting with an AI rather than a human.
To address these concerns, the development team chose the Voiceflow platform and the Claude language model. This combination allows the chatbot to offer controlled responses by providing users with links to library resources rather than attempting to answer questions directly. The system also incorporates disclaimers, user memory, and predefined rules to ensure that the chatbot aligns with the library’s values and operates within ethical boundaries. It is a deliberate design choice meant to avoid the common pitfalls of AI-generated misinformation.
In addition to providing general library support, the team has integrated an AI flow specifically designed to assist users with research. This tool can help users brainstorm topic ideas, generate initial search links, and even provide citation and writing assistance. The AI can also connect users with specialized librarians based on the nature of their inquiry, giving a brief explanation of each librarian’s area of expertise to help guide users toward the most appropriate contact.
While the pilot project has shown promise, the developers are clear about the experimental nature of this tool. While it offers new ways to assist users, there is no immediate push to implement it in its current form. The focus is on exploration—understanding how AI might improve library services in the future, rather than rushing to deploy it as a finished product.
“This project isn’t about replacing human librarians,” explained Choate. “It’s about providing a tool that can fill gaps when staff aren’t available, allowing librarians to focus on more complex and human-centric tasks.”
One of the most significant aspects of this project is the freedom granted by the Libraries to experiment without the immediate pressure of launching a production service. This flexibility has allowed Moutran and Choate to focus on the chatbot’s ethical design and explore the long-term role AI might play in library services.
Presented by Moutran as a Capstone project for the School of Information, the chatbot has received valuable feedback from testers, highlighting the importance of trust and verification of AI-generated information. And her work with the chatbot has provided valuable insights into conversational design, user experience research, and prompt engineering.
As the chatbot continues to develop, the project remains a research initiative with the potential to reshape how libraries use AI. With a focus on transparency, ethics, and user support, the chatbot may one day serve as a supplementary tool for both users and librarians, increasing access to library services and freeing staff to focus on more complex inquiries.
Access the prototype: UT Libraries Assistant Chatbot Demo