AI, research institutions and what comes next: Insights from a leading SN ERAC member

Author: Saskia Hoving

Editor-in-Chief

As artificial intelligence (AI) transforms every part of research (how it’s done, reviewed, and governed), research institutions across Europe are working to understand what these changes mean for institutional policy, governance structures, and long‑term strategy. To explore what these shifts mean at an institutional level, we spoke with Dr. Carles Sierra, Director of the Artificial Intelligence Research Institute (IIIA_CSIC), and member of the Springer Nature European Research Advisory Council (ERAC), about his views on AI’s implications for the research ecosystem.

First, a word about ERAC. To help explore these (and other) issues, Springer Nature’s Research Advisory Councils (RACs) bring together leaders from across the global research ecosystem to explore these challenges. Within this network, ERAC looks at the pressures and opportunities facing European institutions as AI becomes an integral part of research workflows.

We wanted to talk to Dr. Sierra, because he brings a unique outlook on AI in research, as someone who is both an ERAC member as well as an academic AI researcher. With deep experience in both the foundations of AI and its real‑world implications for research and publishing, Carles offers research institutions a valuable perspective on how AI is reshaping science and where institutional leadership and preparation will matter most next.

Human agency in AI-supported research

Dr. Sierra’s role gives him a holistic view of AI in research, and in publishing. He notes that “AI will influence all phases of the research process from hypothesis generation to code generation, to simulation execution, to data collection, and analysis, and even to the writing of scientific papers.” 

From this perspective, Dr. Sierra sees a host of interrelated issues that research institutions need to actively address. The common thread tying these concerns together is the need for institutions to ensure that humans drive the AI, rather than allowing automated systems to shape research practices unchecked.

Essentially, AI should serve science and researchers and making sure that happens requires intentionally centring humans across all phases of the research cycle. Dr. Sierra says, “Human agency must remain at the centre of the process. Decisions should ultimately be taken by humans, and the outputs of AI systems must always be critically evaluated by scientists.”

But keeping humans in charge means that humans need to know how to be in charge. “It is therefore essential to educate researchers about both the possibilities and the limitations of these technologies.” Consequently:

1. Researchers need to know how these models work
2. Researchers and leaders need to be aware of model biases
3. Researchers will need multiple perspectives, and only multidisciplinary teams can give those perspectives

Building AI literacy as an institutional capability

Dr. Sierra says, “Researchers need stronger education and training in AI if they are to adapt their methods… that in many cases have been followed for decades or even centuries… They need to develop at least a good understanding of how these algorithms work and how the AI systems have been trained. It is also important to recognise that some AI systems are not fully reliable and require strong human supervision. AI systems do not have agency, intentions, or moral responsibility.” 

While this applies to researchers at all career stages, institutions play a key role in deciding where, when, and how this education is embedded. For institutions, the best opportunity to build this capability is early, embedding AI literacy into training and development at the start of researchers’ careers. Dr. Sierra said, “Early-career researchers in all fields should acquire a solid understanding of artificial intelligence. It will soon be very difficult to imagine meaningful research in almost any discipline without some familiarity with AI tools and methods.”

Institutional governance and ethical AI leadership

Safeguarding human values in the development and deployment of AI models is increasingly an institutional responsibility. Key questions include: How do we continue to develop and train AI systems and models in ethical ways? How do we make certain we have policies to ensure ethical AI use? These questions actually form the core of Dr. Sierra’s own work, and from that perspective, he says, “We must continue investing not only in the application of AI but also in the science of AI itself… From a scientific perspective, we still need to improve the foundations of AI itself. Important challenges include reducing biases in training data, improving the reasoning capabilities of AI systems, and ensuring that increasingly autonomous systems behave in reliable and ethically acceptable ways.”

From a policy perspective, “Research leaders play a key role in promoting reflection on the responsible use of AI in science. This includes encouraging open discussion, establishing clear institution‑wide guidelines for researchers, investing in AI tools that can accelerate scientific progress, and maintaining close oversight of the ethical implications of these technologies.”

Why multidisciplinary collaboration matters more with AI

Training the models inevitably incorporates various biases into them; not only impacting social biases, but biases that could affect their scientific output. The researchers using these tools need to understand how these biases could affect the output they get from them, to recognise and correct for these biases. Doing this effectively will require institutions to enable collaboration across and between disciplines. Multidisciplinary teams will become more critical, and institutions will need to create the conditions that allow these teams to thrive. Dr. Sierra: “Perhaps the most important competence researchers will need in the future is the ability to work effectively in collaborative teams, combining different types of knowledge and technological expertise… Another important trend is the increasing importance of multidisciplinary teams. Complex scientific problems require collaboration across different areas of expertise, and the boundaries between disciplines are becoming less rigid.”

AI in scholarly publishing, peer review, and evaluation

There are also implications for scientific publishing, spanning everything from writing research to even new ways of publishing it, to research assessment, findability, accessibility, and more. Part of this can involve re-imagining the research article. Dr. Sierra has been collaborating with Springer Nature for several years on the possibility of something called “liquid publications, where scientific documents would have a dynamic digital life. Sections of a paper could be refined, recombined, or adapted for different audiences and purposes.” Well-trained AI could help make a vision like this a reality. 

Beyond that, AI can help lubricate many of the friction points in scientific publishing. Dr. Sierra said, “We discussed the possibility of using AI tools to assist in the assessment of research and how such tools might be integrated into publishers’ workflows. At the same time, we considered what kind of ethical guidelines should be established for the responsible use of AI in the publication process. Overall, there was a shared sense that the way we write, review, and evaluate scientific publications may need to be reconsidered quite urgently.”

Addressing these challenges requires sustained, strategic collaboration between research institutions and publishers, particularly around evaluation practices, incentives, and research integrity frameworks. Dr. Sierra again: “Publishers should listen carefully to the academic community, which is responsible for producing scientific knowledge and advancing research practices. They should observe which practices are emerging and which ones are proving successful. At the same time, publishers should pay attention to the ethical committees and governance structures within academic institutions. Conversely, research institutions also need to understand the pressures and incentives that shape authors’ behaviour, particularly in relation to evaluation and career progression.”

Springer Nature is already addressing some of this by implementing AI tools for research integrity; you can read more about those here.

What to watch next for AI and research support

We’re now at the point where AI’s risks and rewards have come closer, into sharper focus, but the picture is still not quite complete. And as regards AI and research, researchers, librarians, publishers, and funders are all contributing to completing that picture. All of these stakeholders are participating, and they’re all important. 

As research institutions rethink how they approach, use, and govern AI, Springer Nature continues to offer tools, insights, and guidance to support informed institutional decision‑making. You’ll see more perspectives on AI in scholarly communications, including last year’s Perspectives on AI in Scholarly Communications, here on The Link.

About Carles Sierra

Carles Sierra is Director and Research Professor at the Artificial Intelligence Research Institute (IIIA‑CSIC), one of Europe’s leading centres for AI research. He is also an Adjunct Professor at Western Sydney University and a Fellow of EurAI and ELLIS. A prominent figure in multi‑agent systems, agreement technologies, and computational trust, he has authored hundreds of scientific publications, and his work has received 20,000+ citations. Sierra has held major leadership roles, including President of EurAI and Editor‑in‑Chief of the Journal of Autonomous Agents and Multiagent Systems. He has also chaired leading AI conferences such as AAMAS 2009, IJCAI 2011, and IJCAI 2017. His contributions have been recognised with top honours including the ACM/SIGAI Autonomous Agents Research Award and the IJCAI and EurAI Distinguished Service Awards. Sierra also leads AIHUB.CSIC, a network of 400+ AI researchers within CSIC, and works extensively on socially aligned and responsible AI.

Related content

• AI in research and publishing: What institutions need to know from KRAF 2025
• Research integrity in the age of open access and AI: The view from France
• CINCEL libraries explore the opportunities and risks of AI in research support
• Exploring AI with research librarians at Charleston
• Exploring AI futures in libraries with the Futurescape Workshop community

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