Libraries eye multiple uses for artificial intelligence
![](https://www.researchinformation.info/sites/default/files/content/analysis-opinion/lead-image/Screenshot%202024-07-11%20at%2010.03.35.jpeg)
Jon Bentley
AI-driven technologies can serve a vast range of use cases in libraries, explains Jon Bentley
In recent years, artificial intelligence (AI) and machine learning have become buzzwords in technology and media circles, driven by the popularity of chatbots such as ChatGPT.
In many cases, it’s the risks of AI that have dominated headlines. Yet according to speakers and panelists at Access Lab 2024, AI-driven tools can already be applied to a vast range of library use cases.
“Research shows that AI can be like a library superhero, swooping in to make tasks like finding books or recommending resources super-efficient,” said Dr Luba Pirgova-Morgan, research fellow at the University of Leeds in the UK, in a lively opening keynote.
A 2021 survey by the Conference of European National Librarians (CENL) – cited in a report by Dr Pirgova-Morgan – found that European libraries already had AI projects or activities in fields such as layout analysis, text recognition and automatic cataloging, among other uses.
But Dr Pirgova-Morgan was also careful to note that if AI’s potential is to be realised, risks such as privacy and bias still need to be addressed.
The front-end use case
Of course, chatbots such as ChatGPT are often what we think of first when we consider AI. And researchers can already use AI as a front-end tool, to search and discover research literature in useful and powerful ways.
At Access Lab, Bella Ratmelia from Singapore Management University (SMU) libraries introduced delegates to a selection of these.
An initial benefit of AI language models is the way they can answer researchers’ questions in human-friendly syntax. Elicit, for example, generates written responses to queries, which need checking but are supported with citations from the Semantic Scholar database.
Consensus, too, searches Semantic Scholar, while trying to answer yes/no questions; it ranks the “consensus” behind its proposed response. Meanwhile, SciSpace combines a literature review with a range of useful functions such as an APA citation generator.
Other tools help in a different way, by aiming to giving users a “way in” to the mountain of research they confront. Researchers can typically ‘follow the links’ between papers, authors, or topics of study – often while making multidisciplinary leaps in thinking.
Examples include ResearchRabbit, which each help users map such semantic relationships, typically using “mind map” visualisations.
VOSviewer, meanwhile, is used at the University of Leeds as part of its literature search services, according to Looking towards a brighter future, the report presented by Dr Pirgova-Morgan at Access Lab.
Dr Pirgova-Morgan also spoke about the potential of front-end tools to help personalize user search – with appropriate safeguards. And it should also be possible to combine such AI models with publishers’ proprietary datasets or libraries’ own collections.
This may only increase the importance of flexible ways for users to access and interrogate the content at the heart of their research – with seamless access to content remaining critical, so responsible users can double-check the accuracy of AI-generated results.
Content management and digitisation
No front-end system, of course, is useful without an accurate back-end database for it to search. And some AI technologies can help libraries digitize and analyze the arrays of information at their disposal, including offline collections.
CENL’s 2021 survey, for example, found that libraries use a wide range of tools to help digitize and analyze content sources. These include layout analysis, handwritten text recognition (HTR), and optical character recognition (OCR).
At Access Lab, Dr Pirgova-Morgan gave the example of how AI is helping the University of Leeds to digitise ancient texts – with obvious benefits to both researchers and the library.
“They're using AI tools to actually upload images of ancient texts, and pull apart the text from the white spaces, so they can better digitize the text and not have to do it by hand, as it was done in the past. I quite enjoyed looking to see texts that I wouldn't otherwise have been able to decipher myself – seeing an AI tool actually being able to read the complicated letters and writing.”
AI is able to scan other kinds of physical resources, of course. One tool Dr Pirgova-Morgan learned about from a US expert “allows you to scan the base of ships – so that people can train in that virtual space about how to modify and fix the ships, before they go in to actually do it for real”.
Another tool cited in the CENL survey is automated cataloging. This kind of technology could help with the vital process of de-duplicating library records – otherwise a tedious manual job. And Jisc’s Peter Findlay noted that in the US, the library consortium OCLC has also been experimenting with de-duplicating records using AI.
In Looking towards a brighter future, DrPirgova-Morgan notes the potential of AI in managing and organizing collections.
One idea, for example, is to use algorithms to find gaps in collections or to respond to researcher demand – whether that means using tools to suggest new materials, or to find duplicates and work out what to keep.
Further use cases
As part of Access Lab, speakers also covered the potential for AI to make life easier for libraries and learners alike.
“AI tools can be used to streamline library operations, automating mundane tasks, and leaving librarians with more time to focus on more complex and intellectually stimulating aspects of their roles,” Dr Pirgova-Morgan told Access Lab.
The University of Leeds, for example, is exploring using AI to support everyday features such as booking systems for its learning development programs.
Panelist Rosalia da Garcia from Sage Publishing, who studies at Stanford University and uses AI frequently, spoke about the value of a VR classroom and chatbot coaching to enhance the student experience.
In a separate panel discussion on day one of Access Lab, Matthew Weldon of Technology from Sage noted that to encourage AI use among essay-writing students, more clarity is needed on what constitutes plagiarism and what does not. For example, he asked, is it plagiaristic to use AI to write you an essay plan?
Students, too, are getting involved in imagining what AI can look like. SMU Libraries in Singapore, for example, organized a hackathon last summer open to both students and staff, which aimed to “reimagine the future of libraries and research” through AI.
Many of the submissions involved chatbots, SMU’s Bella Ratmelia told Access Lab – with features including tools that could pull recommendations from the library holdings, or pull up the source for a query, which Ratmelia says was innovative at the time.
The future of AI
With so many different possibilities, how can we predict the future of AI in libraries? The answer, perhaps, is that libraries will evolve but remain true to their original purpose.
In her opening keynote, Dr Pirgova-Morgan presented a chapter of research that “highlights the evolving roles of libraries as hybrid spaces that balance [the] physical and digital resources” – an evolution that predates the recent flurry of interest in AI.
But the speech also noted that libraries are still valued as knowledge-focused “sanctuaries”. As long as risks to privacy, accuracy and fairness are managed, there seems little reason why the future might not be bright.
Jon Bentley is commercial director of OpenAthens