Parallel Session 2 – AI and Library Service Development I
Date: Wednesday, 1 July 2026, from 11:00 to 12:30
Moderator: TBC
Location: R7
2.1) Providing Visual Cultural Heritage Material for AI in Research
Presenter: Joona Manner, National Library of Finland, Finland
Science needs AI models trained with high-quality data that also utilise historical data. Usually, this older material is not only challenging to find but also often completely inaccessible or hidden behind extremely difficult-to-use interfaces. However, more and more high-quality visual material is being digitised and becoming available with higher-quality metadata. In addition, traditional research methods are rapidly transforming due to AI, such as RAG (Retrieval-augmented generation) data models.
To address these challenges, the National Library of Finland has developed an extension to the national discovery service Finna.fi. The aim is to provide researchers with novel access to the vast digital cultural heritage materials from almost 500 Finnish libraries, archives, museums, and other organisations. This extension uses Finna.fi’s service API (Application Programming Interface), allowing a user to perform a regular browser-based keyword search and fetch, e.g., a selection or even all high-resolution images from a single institution, using a local Node.js-based solution.
In this presentation, we describe the development of this Finna.fi extension, namely Finna Image and Metadata Download Tool (FIMDT). This tool has recently been tested on a small scale with a limited group of researchers. In the preliminary user tests, we have explored how FIMDT enables downloading a vast number of high-resolution images with corresponding metadata for AI training.
FIMDT has been developed within the FIN-CLARIAH project to provide easier, more efficient access to digital cultural heritage data. FIN-CLARIAH is a research infrastructure for the Social Sciences and Humanities comprising two components: FIN-CLARIN and DARIAH-FI, and part of the pan-European infrastructure for arts and humanities scholars, DARIAH-EU.
Project advancement and the broader transferability of this concept require discussion of the following questions. How to find a consensus and overcome potential obstacles related to the intellectual property rights and AI training? What are the risks and benefits of having free and easy access to such large volumes of cultural heritage materials?
2.2) A Library Led Pilot on Custom GPT Driven Discovery: Strengthening Student Engagement with AI and Information Literacy
Presenters: Cristina Huidiu, Wageningen University & Research, The Netherlands; Alicia Gomez, IE University, Spain
As institutions navigate both the opportunities and challenges of integrating GenAI into learning and research, we believe libraries can play an active role by designing and evaluating tools that promote transparency, critical engagement, and alignment with open knowledge practices.
Within this context, IE University Library in partnership with Wageningen University & Research Library are working on an experimental initiative leveraging IE University’s institutional OpenAI EDU license to build and deploy Custom GPTs for academic information search and analysis. The aim is to explore how generative AI tools, combined with open scholarly infrastructures, in particular with structured scholarly data from the OpenAlex API, can support academic discovery process, with the focus on students’ engagement with scientific literature. Besides, the project aims to better understand how libraries can support the development of robust information literacy skills among both students and academics.
Unlike traditional search systems, this Custom GPT assistant uses natural language dialogue to help users construct and refine Boolean-style search queries, which are then submitted to OpenAlex for retrieval. Importantly, the user is always asked to validate or edit the Boolean query before it is executed, ensuring that human agency and critical thinking remain central to the process. This design places the student in control of retrieval, while allowing the AI to assist with the task of iteratively refining Boolean structures, clearly positioning that AI remains a support tool rather than a decision-maker. Additionally, for open access publications, users can then interact with the scientific content within the same interface.
Students are invited to test the GPT alongside conventional discovery tools, and have to reflect not just on the results retrieved, but also on the process of how generative AI shaped their understanding of keywords, concepts, and citation relevance.
Through these interactions, we collect both performance data and user perceptions via a mixed-method evaluation, in order to gain a comprehensive view of the approach, its value, and its limitations.
This presentation will showcase the development process, student feedback, and key outcomes of the pilot. We will reflect on the implications for AI-assisted discovery systems in academic libraries, the importance of preserving human agency in algorithmic search, and future directions for integrating open scholarly infrastructures like OpenAlex with institutionally governed GenAI tools.
With this experiment, we intend also raise students’ awareness of both, benefits and limitations of the use of GenAI for their work, as the same time that we train students in a very practical way on information literacy in the context of responsible GenAI use. Finally, this study contributes to the broader conversation on how libraries can lead in shaping responsible, transparent, and pedagogically grounded uses of generative AI in higher education.
2.3) Building AI Literacy in the Library World
Presenter: Dagfinn Dybvig, Norwegian University of Science and Technology, Norway
Our proposal emphasizes the critical role libraries are called upon to play in fostering artificial intelligence (AI) literacy amid rapid technological and societal changes. For LIBER 2026, the proposal aligns with the conference theme: The Power of Libraries in an Uncertain World. It argues that research libraries, as central actors in the information ecosystem, are uniquely positioned to safeguard fundamental rights and values while navigating ethical dilemmas and complex decisions in an era of uncertainty.
This, however, requires libraries themselves to work systematically towards increased AI competency within their own organizations. The initiative described—TAIKun, supported by the Norwegian National Library—brings together five institutions in a collaborative effort to develop practical training resources for AI applications in library contexts. The project responds to pressing questions: Why is AI competency essential for libraries? How can libraries ensure responsible and value-driven use of AI tools?
The proposal identifies several thematic areas that libraries must address: geopolitics, cloud exit strategies, secure data storage, ethics, compliance with regulations (such as the EU AI Act and copyright law), and robust user support. These considerations extend beyond traditional library services, like literature searches and systematic reviews. Instead, they encompass advanced AI-powered tools and workflows, including:
- Retrieval-Augmented Generation (RAGs) for controlled knowledge searches
- AI-enhanced platforms such as Scopus AI and Web of Science AI
- Sandbox solutions for tools like Copilot via local infrastructure
- Chatbots for catalog searches and practical library information
- Legal and research tools like Lovdata Pro and Dimensions
- AI integration in screening processes for large amounts of documents
The TAIKun project has structured its work around three core dimensions: ethics, teaching, and research. Constituent groups focus on each of these areas, aiming to create a comprehensive framework for AI education in libraries around a train-the-trainer philosophy. Initial efforts have included mapping existing AI services across institutions and identifyingshared training needs. The next steps involve defining clear learning objectives and developing a flexible, future-oriented training strategy that accommodates rapid technological evolution while remaining anchored in the enduring values of libraries.
Ultimately, the proposal advocates for building resilient, ethically grounded frameworks for AI literacy. By doing so, libraries can lead responsibly in an uncertain world, ensuring equitableaccess to knowledge and supporting informed, critical engagement with AI technologies.