Making AI work for all of scholarly communications — sustainably

Author: Saskia Hoving

Tools leverage and extend human abilities, enhancing our capacity to achieve more. While they come with their own set of challenges, their potential benefits are significant. Artificial intelligence (AI) isn’t really any different. In the scholarly communications world specifically, AI — including but not only Large Language Models (LLMs) like ChatGPT — can help everyone involved with research, including researchers, librarians, and research managers. 

Springer Nature’s recent Sustainable Business Report details many of the approaches that the company is taking to sustainably and ethically develop AI capabilities to serve the entire research community. Some areas where Springer Nature has — and is — developing tools include: Summarising research from across many sources; expanding those summaries into full books; helping journal editors match submissions to the most appropriate reviewers; and more. 

But as with any tool, AI needs guardrails to ensure safety. So even as Springer Nature embeds AI throughout its work, the company also focuses on how to do so ethically and safely, and on keeping the humans in the loop. 

Accelerating Research: Summarizing, speeding peer review, and more

To speed up scientific progress, we focus on what researchers need, helping them work more quickly and efficiently. With AI, we offer a variety of tools and platforms tailored to make their daily tasks easier. These include automated quality and fact checks, AI-assisted copy editing, and translations. This helps improve the quality of their work and allows for faster sharing of scientific findings. By doing this, researchers can concentrate more on what really matters: their research.

1. Helping facilitate peer review and publishing
   To get a sense of the scope and size of research publishing worldwide: Estimates range from 2.14 million new articles (Nature) to over 5 million new articles (WordsRated.com) published each year. And when you multiply that by the number of times the typical article gets rejected and resubmitted, you get an astronomical number of submissions going through peer review. 
   
   A big part of what scholarly publishers do is organise peer review. And AI — by being able to comb through the scholarly record — helps journal editors find and recruit the right reviewers. The efficiency boost that AI tools can bring can both help reduce review times, and also help support open access (OA) and open science more broadly. 
2. Synthesising and summarising current research with machine-generated books 
   Many scholarly books — contributed volumes, especially — are almost like large collections of longer review articles. They collect, synthesise, and make sense of current and recent research, and gather it all in one place. These types of books don’t typically feature primary research or new discoveries.
   
   Collecting, collating, assembling — these are the kinds of tasks that AI can excel at. And when you combine those basic abilities with LLMs, you get a system that can, relatively quickly, generate these kinds of books. Springer Nature has used a system of this sort (with human oversight and editing — of course) to quickly generate a collection of books across a range of topics. This includes titles like  Lithium-Ion Batteries, A Machine-Generated Summary of Current Research, and Climate, Planetary and Evolutionary Sciences: A Machine-Generated Literature Overview.
3. Finding the signal in the noise part I: Text and data mining (TDM)
   The engines driving most of generative AI are called generative pretrained transformers (GPTs). These are the engines — like Open AI’s ChatGPT — driving outputs like text, images, music, and more. And because GPTs work by using past patterns to make content predictions, finding patterns in any dataset fits extremely well with how they work. That makes these systems experts at detecting patterns, trends, and correlations that can be hard for humans to detect. This power has more than one application in research, and one of those applications is in finding research patterns hiding in the literature. 
   
   The process of sorting through vast amounts of data and literature to find these patterns is called text and data mining (TDM). This is a process whereby an AI system can comb through essentially an entire library, and surface connections and insights that it would take humans years to find. 

For librarians: Helping organise information

Publishers can — and arguably, should — use AI tools to better organise the information they publish, to help librarians in curating the collections they develop and hold. Some ways publishers can do this — and some ways Springer Nature is doing this — include:

• Structuring existing content. For example, auto-clustering content (even from across journals or collections) by subject and content. 
• Finding and collecting material from across an entire portfolio, and auto-generating summaries of current research.
• Making recommendations on what to read next, and more. 

All of which can help librarians to better curate their collections, better develop their collections, and, ultimately, better serve their patrons. 

For research managers: Understanding trends

As discussed above in “Part I,” AI’s underlying technology works by learning about patterns, and this includes finding patterns and trends in research outputs — both for the institution generating the research, as well as that institution’s peers. This can show research managers a picture of their institution’s outputs — what disciplines, which journals, what the citation picture is — alongside that of their competitors’. 

Research managers can use these tools to allocate resources, to show funders and administrators the results of their investments, to help write grant proposals, and more. 

Springer Nature’s ethical AI development

As we continue to integrate AI into scholarly communications, it is crucial to remain vigilant about its ethical use. Free AI tools can boost research efficiency and make tasks easier, but we need to be careful with how we use them. This means being clear about how AI works, protecting data privacy, and avoiding biases that could affect research outcomes.

Read Springer Nature’s 2024 Sustainable Business Report for more details on the approaches that the company is taking to sustainably and ethically develop AI capabilities to serve the entire research community.

Related content

• AI in libraries: Improving tradition through innovation
• What to know to plan research training on using AI
• Three ways Generative AI can give Research Managers “superpowers”
• The rise of generative AI: Transforming the role of information professionals
• Can AI help us manage information overload?
• AI and Science Publishing: “Cutting through the clutter has never been more important”
• Text and Data Mining: Uncovering Hidden Data Points and Powering New Discoveries
• Unravelling the Power of TDM: A Transformative Journey for Researchers
• Three ways Generative AI can give Research Managers “superpowers

Don't miss the latest news & blogs, subscribe to The Link Alerts!