IVA's President: Some thoughts on AI’s impact on academic research

Dear Friends of IVA,

We often speak of how AI will transform or even replace professions, occupations and even entire industries. The full impact is yet to come, but one sector where AI is already shaking the foundations is academic research. 

When Open AI announced in September that its AI agents had solved the Navier-Stokes equation, the scientific community immediately began debating the result. But regardless of the final verdict, the announcement prompted Scientific American to declare that “the field will never be the same.”

AI can process huge data sets, design and run experiments, or suggest methodologies – all work that has required and shaped the critical thinking skills and shared values of scientists over centuries. 

It can also write papers. 

Now, an explosion of AI-written scientific papers and funding applications is putting enormous strain on peer review – how we assess and ensure high-quality research. The linchpin of the academic system, this feedback from volunteer academic experts on papers and projects decides what is funded, what is read and (often) who is promoted. These decisions ultimately lead to innovations that improve sustainability, security, education or health outcomes, and keep our companies competitive. AI is now breaking the linchpin of that system, leading some to predict that AI could spell the end of grant funding systems 

This is a particular challenge for the Swedish research system. Sweden allocates more research funding via competitive grants based on peer review (as opposed to basic funding) than many comparable European countries. It also has many public and private research funders, all of which rely heavily on peer review for assessing applications. In response to the latter, a Swedish government commission in 2023 made recommendations for a new agency structure for public R&D funding (FOFIN-utredningen), which, however, were not taken up. 

AI is a stress test for assessment systems like peer review. But, as leading experts in the field have pointed out, the existing system already had significant challenges even before AI entered the picture. Peer review relies on researchers spending considerable time reviewing papers or applications, for which they receive little (if any) compensation or recognition. It can also tend towards conservatism and risk aversion (in what is being researched, how and by whom) and a narrow definition of quality. It is also often painfully slow.

Is Sweden prepared for what is coming? I’m afraid not yet. But perhaps we should welcome AI as an opportunity to reflect collectively and seriously on what universities, research and teaching should be about and redesign our policies and systems accordingly. Such a reflection is overdue, given the massive societal and technological changes of the past decades, as Mats Benner and I argued in, “Change everything!? Towards a higher education policy for our era.” 

This is also why IVA has just launched the initiative “University of the future” under the leadership of Professor Martin Nilsson Jacobi (rector of Chalmers University). I think it is high time to ask how our research assessment and funding system can best serve the advancement of knowledge, and universities’ contribution to society. Then consider new ways of allocating funding – whether focusing on larger projects, creating different kinds of research environments or changing the methods for grant applications. 

In imagining the university of the future, we should not be constrained by constructs that we have come to see as bedrocks of academia. The current pre-publication peer review system is not that old (Nature adopted it only in the 1970s). In the times of Francis Bacon and Isaac Newton, peer review was very different. Scholars published essays or letters and the scientific community reviewed and commented on their ideas – post-publication. Today’s research PhD standard only emerged in the19th century at what is today the Humboldt University in Berlin. China is now experimenting with the concept of a PhD by awarding degrees for “products rather than papers.” 

Why not assess these constructs to ensure they serve science (and thus society) rather than the other way around? As John Douglass, a researcher of universities I greatly respect has said, “Universities are extensions of the societies that give them life and meaning.” Let’s make sure they can serve societies at a moment in human history when we may need them more than ever. Perhaps AI can even help us. This means ensuring it supports quality assessment and funding decisions based on transparency, accountability and excellence, with the aim of creating trustworthy knowledge that serves society.

The research we do at universities matters enormously for whether we can save lives, create economic value, combat and mitigate climate change, protect what is dear to us and create a better future. 

Major changes are coming for research and research funding. It doesn't have to be a bad thing.