When artificial intelligence (AI) can produce confident-sounding answers in an instant, a conversation with a statistician can feel almost sobering. In science, the goal is not certainty or omniscience – it is doubt.
EYSM opening session, Vilnius University Faculty of Mathematics and Informatics
"If you're a researcher, you have to question everything – your own work, even the knowledge," says Dr Elena Bortolato of Pompeu Fabra University and the Data Science Centre at the Barcelona School of Economics.
She was one of the participants at the 25th European Young Statisticians Meeting (EYSM), held in Vilnius from 7 to 10 July. Organised by the Institute of Applied Mathematics at Vilnius University's Faculty of Mathematics and Informatics, the conference brought together early-career researchers from 25 European countries to present work on analysing complex data and tackling real-world problems.
The biennial meeting is one of Europe's leading events for early-career researchers in statistics and probability theory. This year's gathering marked the meeting's return to Lithuania after more than three decades – the eighth EYSM was held in Palanga in 1993. The conference was held under the auspices of the European Regional Committee of the Bernoulli Society for Mathematical Statistics and Probability.
Modern statistics helps us understand and decide
Researchers presented work spanning causal inference, AI and machine learning, fraud detection, forecasting, network analysis, and more robust statistical methods for messy data. Their applications ranged across finance, economics, medicine, genetics, climate science and even nuclear fusion research.
"I'm proud to see this prestigious event return to Lithuania," says Dr Aidas Medžiūnas of the Institute of Applied Mathematics at Vilnius University's Faculty of Mathematics and Informatics, who chaired the Local Organising Committee. "It shows that we have long been – and remain – an important part of Europe's scientific community."
Dr Elena Bortolato
For Dr Bortolato, EYSM is above all a chance to meet colleagues and their work, and to build connections that might one day turn into collaborations. "I like conferences that build community," she says. "It's nice to think we'll keep in touch, maybe even work together in the future."
Professor Jurgita Markevičiūtė, Vice-Dean of the Faculty of Mathematics and Informatics and a member of the organising committee, agrees. "That's what makes EYSM distinctive – the number of participants is limited, and those who want to attend go through a selection process," she says. "What struck me was how well these young researchers could present their ideas, engaging their audience, adapting to it. Many are still working toward their doctorates, yet they're not afraid to question established practice, ask hard questions, and look for answers. Critical thinking is the most important quality a scientist can have."
A discipline that connects the sciences
The programme featured invited talks from internationally recognised researchers in statistics, probability theory, machine learning, stochastic differential equations and applied mathematics.
Among the keynote speakers was Professor Geurt Jongbloed of Delft University of Technology's Institute of Applied Mathematics, whose research develops new methods for making sense of complex, messy or incomplete data – work with applications in biology, medicine, engineering and climate science.
Prof. Jongbloed first visited Lithuania in 1993, for the eighth EYSM in Palanga. He still holds warm memories of that meeting in a newly independent Lithuania, the Lithuanian words he picked up, the friendships he made, a visit to Nobel laureate Thomas Mann's summer house in Nida.
Prof. Jongbloed brought back to Vilnius the same small Lithuanian tricolour pin that his colleague Prof. Remigijus Leipus from the Faculty of Mathematics and Informatics had given him back then.
More than numbers and charts
Say the word "statistics" and most people picture tables, averages, charts. For Prof. Jongbloed, descriptive statistics is only one small part of the field. Mathematical statistics goes further – it's about building the mathematical models and methods that let us use those models for forecasting, spotting trends in data, or comparing outcomes, such as the differing effects of competing medical treatments.
Prof. Geurt Jongbloed
"A big part of our work is understanding the strengths and weaknesses of widely used methods, and the assumptions behind them," Prof. Jongbloed says. "New types of data and new questions keep appearing. So it matters not just to understand and correctly apply existing methods, but to build new models and new ways of analysing them."
Statistics in an anxious society
Anxiety about deep uncertainty – over climate change, health, politics, education, artificial intelligence – is a recurring theme in public debate. Statisticians can end up cast as latter-day oracles – people who understand the data of the past and present, and so should be able to offer precise forecasts, reduce uncertainty and free society from its chronic anxiety.
Prof. Jongbloed sees it differently. Science's goal, he argues, isn't to eliminate uncertainty – it's to understand it. "It matters to grasp not just the uncertainty itself, but the nature of the models that our conclusions rest on," he says.
Scientific conclusions are built on data, he notes, and real data always carries some degree of uncertainty. Presenting results as fixed, unchanging truth would be misleading. Understanding how much uncertainty exists, where it comes from and how it shapes conclusions is a core part of interpreting research honestly.
Prof. Jongbloed offers an example: average winter temperatures in Vilnius. Ask whether the average is rising, and traditional statistical methods would fit a single "best" line through the data points and check whether it trends upward. A shape-constrained approach – one that imposes looser conditions on the form a curve can take – allows more flexibility. "Comparing that curve with the straight line lets you judge more precisely how well-founded the claim really is – that Vilnius winters have genuinely warmed over the period in question," Prof. Jongbloed says.
Decisions built on data
In recent years, "data-driven decision-making" and "evidence-based policy" have become fixtures of public policy language – the idea that decisions and their implementation should rest not only on political ideology, election platforms, public reaction or lobbying, but on objective data, research and systematic analysis.
Yet the phrase can mean different things to different people.
What counts as "evidence"?
Where's the line between grounding a decision in data and letting data dictate it – stripping away the accountability of the people or political groups actually responsible?
And how much data is enough to justify a decision, when some aspects of reality resist being reduced to numbers at all?
For Prof. Jongbloed, what matters is that decisions are made with as much relevant information as possible, while recognising the limits of that information – that meaningful data may simply be unavailable. "In the end, important decisions should always be made by people, who weigh up their sources of information and the possible consequences of their choices," he says.
AI is changing universities
That same responsible approach to weighing sources of information increasingly has to extend to new technologies too. Data is accumulating faster than ever, and large language models (LLMs) — commonly known as AI – now let people access information more easily and more quickly than before. Could AI help make public debate more evidence-based?
"AI is fundamentally built on statistics, and it can become a valuable tool," Prof. Jongbloed says. "But it has to be used responsibly."
Panel discussion on AI impact at VU MIF
AI's impact on science and society was also the subject of EYSM's one open public event – a panel discussion moderated by Dr Medžiūnas. Researchers shared how AI saves time for them and their students, even as it continues to make mistakes and requires careful human oversight.
Dr Balázs Csanád Csáji (HUN-REN and Eötvös Loránd University, Hungary) observed that researchers sometimes treat AI like a lottery machine for generating ideas and argued that what matters is using it to sharpen your own thinking, not letting it replace that thinking altogether.
The panel ranged widely – how AI is reshaping trust in science, the difference between results and proof, and broader economic and social questions, including unequal access to advanced AI tools across countries and universities.
Dr Xiaocheng Shang, of the University of Birmingham in the UK, argued that AI's value in research has less to do with subscription cost than with user skill: "What matters more is how you train the model, how well you write your prompts."
Dr Carlos Escudero Liébana (UNED, Spain) suggested leaving the environmental cost of the data centres AI depends on to market forces – electricity and water pricing – and called for closer collaboration between universities and industry.
Dr Marta González García
Dr Marta González García (Valencian International University, Spain) countered that understanding what's genuinely worthwhile in a market requires factoring in external costs too – like impacts on the environment and human health.
Dr González García also raised the question of "deskilling" – how easily students today can produce a "correct" piece of work without ever learning the underlying concepts, without effort, without correcting their own mistakes.
Prof. Jongbloed shares that concern. Young researchers, he suggests, may find it harder than previous generations to engage deeply with their subject, given AI's growing pull. Looking ahead, he urges young scientists to choose problems that genuinely interest them, and to take the time for real depth of understanding – building not just technical skill, but scientific intuition.
10 July, 2026