Data-SKUP 2026

Som datajournalist kan det noen ganger føles som om språkmodeller og andre AI-verktøy ikke fungerer særlig godt til oppgavene man jobber med. Det holder ikke at ett enkelt eksempel blir riktig – alle dataene må være korrekte. Å bruke kode for å analysere data er dessuten noe annet enn å kode en widget til nettsiden. Helena Bengtsson viser eksempler fra arbeidet sitt med lokalaviser i Sverige og gir tips om hva som fungerer – og hva som ikke fungerer. Hvordan bør man tenke når man bruker AI som datajournalist?

Hver uke produserer norske kommunestyrer tusenvis av sider med sakspapirer. Demokratibasen samler dokumentene automatisk og bruker KI til å oppsummere sakene, trekke ut navn og selskaper, og flagge det som kan være en nyhet. I denne sesjonen får du se hvordan du kan bruke verktøyet i jakt på politiske saker, hvordan det ble til og hva de har lært om å bygge KI-verktøy for journalistikk.

From a farmer in the wet West Flanders who sees the end of the profitable vegetables business he rules to the hundreds of villages that can not drink anymore the water from their taps in Spain or Greece. From the damage to the wetlands in Italy to the 1,300 wells oblied to be close in France and the 300 types of different pollutants found in their water. Nitrates, PFAS, fuels, medical components, overexplotaition. In the worst moment, when climate change and harder droughts already dry huge surface water resources, the europeans, so sure they had the privilege of unlimited access to drinkable, high quality water, are about to discover a terrible reality. Come to know how we made it.

How much protected land and sea have we lost globally to oil and gas extraction? To answer this question, 13 media outlets across four continents embarked on a mapping exercise of epic proportions. The "Fueling Ecocide" — a cross-border, collaborative data investigation coordinated by EIC (European Investigative Collaborations) and EIF (Environmental Investigative Forum) — compared 315,000 protected areas from all over the world with 15,000 extraction blocks, spanning 120 countries. This session will take you through the makings of this data-driven global investigation. Reporters will talk you through the trial-and-error process of handling and harmonizing a large geospatial dataset: which tools they used, which methodological choices they made, which obstacles proved the hardest to overcome, and what lessons they learned along the way.

Prior to the 1990s Europe’s landfill sites were simply unlined holes in the ground where waste was dumped, buried and forgotten about. For some parts of Europe, they have been completely forgotten and data is scarce. Of those we do know the location, many are located in flood-risk zones, areas used for drinking water, or in sensitive conservation sites. Many could be releasing toxic waste into the environment around them. In this session, Juliet will look at the problems of gathering data and how to approach a cross-border story where data doesn’t give an accurate Europe-wide picture. She will discuss how to develop a methodology and how to use different datasets to find a story.