Parallel Session 10 – Artificial Intelligence for/in/and Research Libraries III
Date: Thursday, 3 July 2025, from 09:00 to 10:30
Moderator: Dagnija Baltina, National Library of Latvia, Latvia
Location: Room 2106
10.1) The Role of Research Libraries in Supporting Responsible Generative AI Use
Presenter: Johan Philips, KU Leuven, Belgium
The rapidly evolving and wide availability of Generative AI (GenAI) tools raises questions regarding its use for academic education and research. To address the innovative nature and crosscutting impact of this technology, our university formed a GenAI expert panel as advisory body with interdisciplinary experts across academic staff and central services. Promoting responsible GenAI use to students, researchers and staff was identified as the essential first step to address opportunities, challenges, and risks related to the growing wave of GenAI tools.
This presentation aims to convey two messages from our lessons learned in deploying our university wide GenAI support: (1) Research libraries play a crucial role in providing GenAI advice, offering education and bespoke training. Indeed, their staff is skilled in information literacy (e.g. information retrieval, evaluation and referencing) and, thus, is well placed to assess the quality of various GenAI tools in an academic context; (2) A coordinated approach is necessary, from the start, with well defined responsibilities for all involved stakeholders, including libraries, educational services, and the research coordination office.
To improve responsible GenAI use, we opted for a dual solution with a tiered support system for questions and a set of thematic courses. By intensifying internal collaboration early in the process, we leverage interdisciplinary expertise to address the broad application and impact of GenAI.
Several support tiers were created to complement the direct line of support (zeroth tier) consisting of the immediate environment of students, researchers and lecturers.
The first tier is considered the central point of contact and answers generic questions on GenAI and responsible use without needing domain specific knowledge.
The second tier answers more specific questions such as specific tool use for information retrieval, but also in other fields such as research data management or scholarly publishing.
The third tier typically handles technical, legal or policy related issues regarding GenAI use, that are very specific or complex and, sometimes, require advice from the GenAI expert panel.
The development of course material and efforts in educating students and researchers in responsible GenAI use complements the tiered support. Libraries should coordinate here with relevant partners as well, typically educational services and the research coordination office. Our course material is designed to teach students and researchers the general concepts of GenAI literacy and provide them with criteria to critically evaluate tools and methods.
In conclusion, the integration of GenAI tools in academic settings necessitates a comprehensive and collaborative approach to ensure responsible use. Libraries are pivotal in guiding students and researchers through these technologies, leveraging their expertise in information literacy. A tiered support system, combined with thematic courses, provides a robust framework for addressing the diverse needs and challenges posed by GenAI. By fostering early and ongoing collaboration among all stakeholders, universities can effectively harness the potential of GenAI while mitigating associated risks. This coordinated effort not only enhances AI literacy but also ensures that the academic community is well-equipped to navigate the evolving landscape of GenAI responsibly.
10.2) Scan a Book to Change your Future: The TU Delft Library’s Holo-vitrine Ai Librarian Prototype
Presenters: Vincent Cellucci and Alice Motta Maia Bodanzky, TU Delft, Netherlands
University libraries are evolving beyond traditional roles as repositories of books, emerging as innovative spaces that inspire learning and engagement. In this context, we present the AI Librarian Holo-vitrine, a prototype developed as part of the TU Delft Library’s ongoing transformation of its book wall into a dynamic Collection Wall aimed at visualizing and enhancing user interaction with both physical and digital collections. This interactive holographic system integrates artificial intelligence (AI) and RFID technology to captivate students with speculative career predictions based on scanned book data. Positioned between the shelves, it works in combination with a new book recommendation system, where expert educators highlight recent titles.
Designed to address barriers to student engagement, this system offers a playful and low-threshold method for exploring library resources. Through scanning books, students receive speculative job titles and career descriptions based on the book’s content. The system’s chatbot avatar—formed from a particle cloud of books—engages users in a futuristic, whimsical interaction, sparking curiosity and making library collections more approachable and less intimidating than traditional catalog systems.
Developed in collaboration with the Library’s New Media Centre and Studium Generale, the design process involved iterative prototyping and extensive student feedback to ensure usability and relevance. Early responses have been promising, suggesting that this integration of speculative storytelling, interactive technology, and physical collections has significant potential in reshaping how libraries engage with users and inspire collections exploration. The AI Librarian, like all librarians, emphasizes the transformative power of reading in shaping future career paths.
Video of the experience: https://www.youtube.com/watch?v=ZhuYLphfneM
10.3) Using AI to Transform Old Card Catalogs and Increasing the Visibility of Collections
Presenters: Karin Byström, Uppsala University Library, Sweden; Åsa Forsberg, Lund University Library, Sweden; Markus Nolgren, Göteborg University Library, Sweden
This presentation will give a practical example of how libraries can use AI and modern tools to transform old card catalogs to access in online catalogs, and thereby enhancing the visibility of older collections. By large-scale automated cataloging efforts, libraries support both their users and lay the foundation for stronger library collaborations on collection development.
At the three Swedish University Libraries in Uppsala, Lund and Gothenburg, ongoing projects are focused on enhancing the visibility of older materials currently only searchable through handwritten or typewritten card catalogs. The projects employ AI models to transcribe and decode metadata from these catalogs, and use APIs to automatically generate holding information in the national catalog Libris.
The presentation will outline the methodologies and challenges faced by the projects, regarding choosing and training of the AI models, extracting metadata from the transcribed text, mapping the metadata to records in the national catalog and creating holding records. In addition to technical aspects, the presentation will also discuss the collaborative opportunities of the projects, both between the three university libraries and with the National Library, and the different types of skills needed.
All three libraries have large print collections that are only catalogued in handwritten or typewritten catalogs, going up until the 1960’s, with millions of copies among the three libraries. By the current projects, they have acted on the need to improve the accessibility of the materials in these catalogs. Despite being scanned and readable online, the difficulties of using the older card catalogs have long been a major hinder of discoverability, making the material underused and almost invisible since it is not searchable in the national catalog.
The effort of making collections visible in online catalogs is also important for national and international library collaborations. This has become clear in the context of Swedish national collaborations on digitization and shared responsibility for long-term access to print library material, where bibliographic and holdings metadata must be able to be shared in a digital format.