Presenter: Alicia Lillich, NIH

This lightning talk will focus on the development and delivery of a library instructional session designed to help researchers and teams build a practical, tailored framework for integrating generative AI tools into their work. The class guides participants through a process that evaluates how AI tools fit into their specific research needs, helps them develop a custom strategy for AI implementation, and emphasizes the importance of documenting their approach to maintain research integrity and reproducibility.

Presenter could not join

Presenter: Hilary Murusmith, Princeton University Library 

When working with a book or serial published in a foreign language, catalogers may sometimes be the first to create a record in their language of cataloging, although there may be an excellent MARC record in another language from which a new record can be derived. However, deriving a new record can be a tedious task that does not require further judgment and expertise.

As a member of an internal interest group on generative AI in cataloging, I’ve been exploring if Copilot can help with this particular task. In my talk, I will go over how I developed an effective prompt and the steps I followed to export records from OCLC and include them in my prompt. Then, I will show my experimentation with five non-English language MARC records and how I was able to refine my prompt based on the results I received. I will conclude with an evaluation of the results and considerations like how to establish a workflow that would result in time savings. While this talk covers a niche application of generative AI in cataloging, I hope my experience will encourage catalogers to try using generative AI to address specific problem areas in their work.

Presenter: Mingyan Li, Kavita Mundle, Ian Collins, University of Illinois Chicago. Eric H. C. Chow.

Libraries worldwide are exploring the integration of Large Language Models (LLM), generative AI, and machine learning to enhance metadata workflows. This research was conducted by librarians from the University of Illinois Chicago and Hong Kong Baptist University, focusing on using Google Gemini and OpenAI to generate descriptions and subjects for videos of silent films. Several Google Gemini prompts have been tested to guide AI tools in developing the most accurate results. In the next step of this research project, different implementations for the AI-assisted metadata workflow will be tested and compared to identify the best timing for human intervention. This research will help gain valuable insight on AI's potential and challenges in providing and enhancing resource description and discovery.

Presenter: Sonia Yaco, Rutgers University

This session presents the results of a multi-year research project that identified AI tools that can improve access and enhance discovery capabilities in archival collections, allowing comparisons across disparate materials. The project is a collaboration with Rutgers and Durham University UK cultural heritage professionals, humanities scholars, and computer science experts. Thirty software tools were applied to untranscribed handwritten and typewritten documents and photographs in an English language Japanese cultural heritage collection at Rutgers University Libraries. The findings are presented as three use cases, showing that AI can mine text, find insights in collections, and link text to photographs. Scenario One discusses accessibility through transcription, translation, and narrative extraction. Scenario Two introduces tools for data pattern exploration to provide deeper understanding of collection content. Scenario Three shows ways to improve discoverability and integration of multimedia collections. The purpose of the session is to provide an understanding of practical ways that AI can be used with distinctive collections. Barriers to the use of AI will be discussed including ethical considerations, the need for infrastructure support, and policy challenges. The authors’ experiences with AI tools and implementation will provide a useful guide to others looking to explore AI for their archival collections.

Presenter: Aaron Pahl, University of Alabama Birmingham

AI has the potential to transform how we approach archiving, research, and daily library operations.

I will demonstrate how AI tools can be utilized to enhance efficiency and creativity in the field. I will explore how AI can assist with generating scripts and organizing ideas. The emergence of AI chatbots opened possibilities for cultural heritage practitioners who often wear many hats that require finding innovative solutions to problems outside of their area of expertise. We will explore how to generate scripts from text prompts using AI.

At a basic level librarians can leverage AI chatbots as a starting point to organize jumbled thoughts into a coherent draft before ultimately rewriting the material. This is especially beneficial when staring at a blank page looking for that first sentence.

I will also highlight the experimental use of AI in creating virtual 3D exhibits, which has the potential to create engaging and interactive displays that bring archival materials to life, offering new ways to interact with collections.

By showcasing practical examples and case studies, I will provide attendees with actionable insights and tools to implement AI in their own workflows. The goal is to inspire librarians to harness AI in their professional roles.

Presenter: Nancy Richey, Western Kentucky University

Special Collections Libraries and archives are being asked about the use of AI and Genealogical Research. This form of research remains at the top of how patrons continue to come into the library for traditional research. How is AI being used in family history research and what are the problems and promises for both patron and the library.

Presenter(s): Max Prud’homme, Jeevithesh Cattamanchi Venu, and Hiranya Pappu, Oklahoma State University 

The Oklahoma State University Digital Archives are in a new phase of improving the discovery and preservation of historical materials by leveraging advanced digital methods, such as facial recognition and machine learning technologies. Facing challenges like sparse and inconsistent descriptive metadata, decreasing resources, and a fading institutional memory, the Archives seek to enhance archival asset discovery and optimize computational processes to enhance access and preservation to cultural heritage materials. To achieve this, Digital Archives are integrating AI technologies, applying a deep learning framework designed for face analysis and recognition. With consideration to scalability and sustainability, the team has developed a universal model to enhance descriptive metadata, search functionality, and discovery. The development team coordinates efforts with the metadata team that organizes, creates, and edits thousands of archival records. The objective is to augment the value and improve the searchability and visibility of digital materials to better contextualize the whole university collection of faculty, alumni, as well as buildings and more since its founding in 1890. The presenters propose to showcase the project flow, context, planning, design and architecture, focusing on scalability, sustainability, and ethical issues associated with face recognition technology.

Presenter: Amy James, Baylor University

In today’s rapidly evolving digital landscape, AI tools are becoming commonplace for student researchers. However, the convenience of generative AI often comes with significant challenges, particularly in the realm of information accuracy and ethics. As a librarian at Baylor University, I identified a growing concern among students in our EdD Learning and Organizational Change (EdD LOC) program: students were using generative AI, but stopping with the initial response(s) that it provided, i.e. they were not critically evaluating the AI-generated content. This not only compromised the quality of their work but also posed ethical dilemmas, as they were citing hallucinated articles in their dissertations.

In response, I developed a targeted lesson aimed at enhancing AI literacy among our EdD LOC students. The session focused on going beyond the initial responses provided by generative AI in research, emphasizing the importance of verifying AI-provided sources. I also introduced practical strategies for integrating AI responsibly into their academic work.

This lightning talk will highlight the key components of the lesson, share insights into the challenges and successes encountered, and provide actionable steps for other librarians looking to implement similar initiatives. By fostering a deeper understanding of how students are using generative AI in academia, we can better prepare our students to navigate the complexities of AI in research.

Presenter: Tomeka Jackson, Jen M. Jones. Clemson University 

AI can assist with creating more specific subject headings and can be used to generate summaries for films. The use of AI, in this instance, allows for more accurate subject heading assignments. However, human catalogers should still review AI-generated suggestions. Ultimately, using AI in conjunction with LCSH can improve the discoverability of moving images.

Clemson University's Metadata Services Team has experimented with AI Tools to assist with cataloging and metadata workflows. Recently, we used Microsoft Copilot to evaluate the use of Library of Congress Subject headings and Genre forms. In this presentation, we will show the process of this experiment using screenshots and examples to discuss the advantages and disadvantages these tools create in cataloging various materials and how AI tools can help assist with effective workflows .

Presenter: Jeremy Nelson, Stanford University Libraries

A major benefit of the open-source FOLIO Library Services Platform is the community's ability to experiment with new technologies and features. This presentation will demonstrate a proof-of-concept back-end module called "edge-ai" that uses Large Language Models (LLMs) to automate the creation of FOLIO JSON records from textual prompts and uploaded images of books. As part of the demonstration, we will address privacy concerns, hallucinations, and other challenges associated with using fast-changing LLMs in the edge-ai module. Additionally, we'll introduce the iterative development process of the edge-ai module as we explore support for other generative AI use cases in circulation, finance, and various FOLIO functions for libraries. Presentation website is available at: https://jermnelson.github.io/ai4libraries-2024/

Presenter: Meg McMahon, Chris Hyde, Marc McGee, and Betts Coup, Harvard Library

This presentation will explore how Harvard Library structured a project to examine applications of conversational AI to support work and better serve their community. Fourteen library staff representing different departments and library work participated in this project using ChatGPT+ to conduct various structured experiments relevant to library work. Within the presentation, the team will explain how they conceptualized and built a community of practice engagement model for this work, along with the deliverables of a library-specific use case library and prompting guidance.

After discussing the project setup, three library staff members will showcase library-specific use cases related to technical services and archival work. First, there will be a demo of using ChatGPT+ to help create usable finding aid content from dealer lists of archival collections. Second, there will be a demo of using ChatGPT+ to create Python scripts for pulling files out of downloaded objects from a preservation system. Finally, there will be a demo of using ChatGPT+ to help make Python scripts for unique ID associations within geojson files.

By sharing the group’s experiences, we aim to inspire conference attendees to think of new ways to structure staff engagement and experimentation with generative AI tools.

Helpful Links:

• Prompt writing examples for ChatGPT

• ChatGPT+ Use Cases

Presenter(s): Richard Song, Qin Lippert, Epsilla.

Join us for a practical, hands-on session where you'll explore the transformative power of AI and Retrieval-Augmented Generation (RAG) to enhance your library's content and services. We’ll start with a high-level overview of AI technologies, followed by demonstrations of practical AI applications in libraries, including AI-powered research guides, FAQs, and special collections management. You'll then have the opportunity to create your own AI-driven applications, such as chatbots or smart search tools, tailored specifically to your library's needs. Whether you're aiming to enhance user experience or improve content discovery, this session will equip you with the knowledge and tools to implement AI solutions in your library. Come prepared to get hands-on and leave with actionable skills to elevate your library's digital offerings.