An “Aye” for AI? Librarians in the Age of AI

Artificial intelligence has the potential to impact many functions of the library. Are librarians ready to seize the opportunities?

What comes to mind when you consider the functions of the library facing the greatest impact from Artificial Intelligence (AI)? Participants at the “Navigating the AI Implementation at ASEAN Libraries” webinar held on 24 April 2024 gave the following responses:

Figure 1. Word cloud of participant responses to the question “Which functions of the library are most impacted by AI?”

So, pretty much every aspect of the libraries then.

Clearly, AI has the potential to impact libraries across many aspects of the information research workflow, such as searching and research, library reference services, circulation, and cataloguing, among others. It is with this understanding of AI’s pervasive impact across all library services and functions that the National University of Singapore (NUS) Libraries team explored how they navigate this uncharted territory as an academic library at a research-intensive institution. Titled “An ‘Aye’ for AI?”, the NUS Libraries team approached the topic by addressing four key questions pertaining to AI and librarianship.

Question 1: How can librarians navigate information literacy in an era of AI while staying true to our values and heritage?

Librarianship has always been instrumental in leveraging emerging technologies to enable access to knowledge, adapting ancient methods of communication and information storage such as cuneiform and papyrus, for contemporary audiences. Libraries have accordingly evolved alongside changes in communication channels and information formats, ensuring continued relevance and accessibility. Looking back across the span of history, we observe that while methods and formats have constantly evolved, the fundamental roles and functions of librarians remain constant.

Accordingly, as a complement to this process of technological change, librarians have led information literacy education through the years. At NUS Libraries, this is codified and articulated through our Research Skills Framework (see Figure 2, below), which sets out fundamental information research skills such as ideating, searching, using and citing, as well as evaluating. Our ambit has also recently expanded into information management roles such as research data management, managing our institutional repository, and archival.

Figure 2. NUS integrates AI tools and information literacy issues relating to AI into existing courses which teach skills related to domains in our Research Skills Framework

Amidst the hype-fear narrative that often surrounds the impact of AI on various professions, the question we need to ask ourselves is not whether AI will render librarians’ skills obsolete, but what new forms these skills will take in an age of AI. AI or not, users are still going to search for, evaluate, cite and manage information. The main challenge for librarians will be understanding how to apply information literacy to an information landscape increasingly dominated by AI and algorithms to integrate in every part of our library research skills framework.

This is especially so for academic libraries, who must align with wider institutional guidelines and policies towards embracing AI and respond accordingly. For example, when ChatGPT came to the forefront of the world’s attention in November 2022, NUS and NUS Libraries’ plans echoed Singapore’s national stance in wanting to be progressive in embracing AI. The University convened a University Policy Workgroup for AI in Teaching and Learning, which sets out guidelines for how the University should use AI. The library aligned with these guidelines when developing our programmes.

Question 2: What skills should librarians build-in to adapt in an age of AI?

Given the rapid development and continually evolving impact of AI on libraries, it will not be possible to provide an exhaustive list of skills and competencies to develop. However, NUS Libraries has identified some key skills and competencies that we’ve started developing in our teams, to enable them to adapt to AI.

For librarians that deal with technical aspects of library operations such as institutional repositories and library IT systems, technical skills and know-how like machine learning and mathematical techniques and useful software libraries are relevant. These can be incorporated as part of systems management and IT skills to introduce and sustain new AI-based library services.

As for instructional librarians, coding or programming knowledge may not be essential but an understanding of AI interfaces and evaluation considerations may be needed to translate existing professional skills like information literacy to an AI world. One such situation could be the use of generative AI tools to create ideas for exploring a research topic and uncover research gaps in the literature. Generative AI tools to can also be used to come up with search strategies and synonyms for database searches. Another application could be using AI tools to compute the relevance of articles for the purpose of screening, as well as evaluating the outputs or results from algorithms.

However, acquiring skills are not enough. We must also think about how to adapt pedagogy and content in terms of learning outcomes, content and delivery methods to effectively raise awareness in our library communities.

At NUS Libraries specifically, we conduct course-integrated programmes, and library-initiated workshops that reach out to students and the academic community of all levels. We also collaborate with academic libraries across Singapore to organize events promoting knowledge sharing on AI.

Question 3: How can we apply AI technology in library services and operations?

There are many positive ways, big and small, of applying AI technology in library services and operations. On one end of the spectrum, we can build custom solutions that enhance discovery and use of our collections. On the other end of the spectrum, we can incorporate commercial off-the-shelf AI models to make certain parts of our work easier.

One example of a custom-built AI solution is the NUS Libraries article recommender, which we built in collaboration with AI Singapore in 2020, and was in service until early 2024. Our goal is to recommend relevant academic articles to users based on similarity of their interests to other users and content. To this end, we make use of usage data for e-resource and physical items to train two AI models based on content-based and collaborative filtering. The AI algorithm then comes up with suitable recommendations when users search for articles, or when they opt to receive weekly recommendation emails.

Figure 3. Mapping subject clusters to Library of Congress Subject Headings using machine learning

Those of us with tighter budgets can consider leveraging freely available AI tools to enhance library operations and work. For example, we can analyse a collection or circulation logs by clustering similar items. We can do so by “vectorizing” selected text such as title or subject headings (see Figure 3, above), converting words, sentences or passages into numbers (or vectors) based on the relationship of one word with another in terms of positions of neighbouring words and passages. We then take this vectorized data and apply a clustering algorithm, such as DBSCAN, which assigns the keywords into clusters based on characteristics such as density and minimum number of data points. Such vectorization and clustering packages are freely available and can be installed as part of R or Python libraries. However, it must be noted that freely available AI tools may lack the support and user-friendliness of commercially available AI tools so additional resources may be required to employ free tools effectively.

Question 4: How do librarians acquire the necessary skills to be effective information professionals that both incorporate AI as part of the information literacy landscape, and leverages AI in their work?

It can be overwhelming to be confronted with a technological paradigm that is constantly evolving! However, we should take a step back and reflect on the progression of librarianship over the years. Those of us of a certain vintage would have already seen and experienced the many changes in library technology and service delivery that have taken place over the years. Many tools and services that we use and offer now (e.g. data services and digital humanities) were revolutionary in their time, but are now considered commonplace. What has not changed is the librarian’s fundamental role and raison d’etre – to be the flag bearer and primary advocate for knowledge management, and for those who seek to access that knowledge for the betterment of themselves and society at large. That should give us the focus and confidence to harness AI in advancing library services for the libraries of the future.

To do so effectively, we need to establish AI technology as part of our essential understanding and knowledge as librarians. While we need not be proficient in creating or even using AI solutions, we should all have a baseline understanding of the family of AI techniques, how they are applied in tools and solutions, how best to leverage these solutions, and having an awareness of potential pitfalls such as cognitive biases and other ethical issues.

Data literacy is a good place to start our upskilling since so much of AI relies on shaping and modeling data. There are numerous such programmes available in the market, such as through online learning. NUS has created a mandatory Data Literacy Programme (DLP) and AI Competency Course (AICC), to give all NUS staff a basic level of awareness and competency in these fields. Intermediate and advanced-level courses are also available for those who wish to learn more about working with data and leveraging AI in their work.

While formal training plays a part in upskilling librarians, informal and non-formal learning methods such as learning on-the-job, or hands-on experimentation can also be very effective in helping librarians gain competency and confidence in using AI. Nothing beats having an immersive hands-on experience to develop tacit, practical knowledge that expands upon skills taught in formal courses!

Besides an understanding of data science and AI, it is also important for librarians to be able to make effective decisions on application of AI in information workflow, implement the solutions, and communicate effectively with users. When we asked webinar participants “What are your plans for learning about AI in the Library?”, many specified areas they wished to integrate AI into, but did not specify how they planned on upskilling themselves. This indicates that generative AI technology is still in a nascent stage, and participants’ library communities have not yet incorporated them as part of the information landscape.

Conclusion*

AI is a dynamic and rapidly evolving field that can transform the library and information services. To effectively leverage AI’s capabilities and provide innovative solutions, librarians must equip themselves with the relevant skills and knowledge. However, formal training alone is not sufficient to achieve this goal. To embrace AI in entirety, librarians also need to engage in informal, non-formal learning and reflective methods that allow them to experiment the AI tools and applications, learn from their peers and experts. In doing so, we can develop the confidence and competence to integrate AI into our workflow and enhance our value as information professionals. This adaptive learning mindset will be crucial for librarians to stay ahead of the curve and maximize the potential of AI enhancing library services and user experiences.

*Conclusion powered by Microsoft Copilot. The rest of the paper was drafted without AI assistance 😊

Contributed by:

Magdeline NG Tao Tao, Vice University Librarian, NUS Libraries
Eng Aun CHENG, Principal Librarian (Library Data Analytics & Procurement), NUS Libraries
Siu Chen LIM, Senior Librarian (Research Librarian – HASS), NUS Libraries
Edited by Marcus WONG, Associate Director (Marcomms & Business Development Cluster), NUS Libraries

Acknowledgements:

The authors would also like to thank all who contributed to the presentation in one way or another including the NUS Marcomms Team and Lee Cheng Ean. NUS Librarians whose work was featured in this presentation: Annelissa Chin, Patsy Chia, Raudhah Thongkam, Chris Tang, Diyana, Lyndia Chen, Mak Jie Ying, Tan Poh Lay, Winnifred Wong, NUS Libraries Copyright Team. Organizations, Associations and Alliances: Singapore Alliance of University Libraries, NUS Information Technology Department, Library Association of Singapore, National Library of Singapore and AI Singapore.