SLU researcher receives ERC Advanced Grant to use AI to stress-test conservation policy before implementation

Guillaume Chapron, a researcher at the Department of Ecology, has been awarded2.5 [GC1] million Euro by the European Research Council (ERC), the EU’s leading funding body for frontier research. SLU has previously received ERC funding, but Guillaume Chapron is the first researcher at the university to be awarded an ERC Advanced Grant. The grant is aimed at established research leaders seeking to carry out ambitious, ground-breaking projects. For the ERC 2025 Advanced call, 3240 proposals were evaluated and 317 were selected for funding across all disciplines and in all EU countries.

“The goal is to create a general method that makes it possible to stress-test conservation policies before they are put into practice. Thanks to rapid advances in artificial intelligence, it is now possible to build models that better capture human behaviour, values, conflicts and political processes,” says Guillaume Chapron.

Failing to protect biodiversity

Biodiversity continues to decline despite international agreements and species protection measures. Failures may sometimes result from insufficient knowledge of ecology, but more often they stem from the ways in which people are affected by, and respond to, conservation measures. Such measures may create social conflicts, lack local support, face organised opposition from different interest groups or even be politically exploited for reasons unrelated to conservation.

“There are sophisticated models for predicting complex ecological processes, but far fewer tools capable of identifying when conservation measures risk trigger socio-political negative feedback that may undermine conservation goals,” says Guillaume Chapron.

Creating stakeholders with AI

The project will use large language models (LLMs) to simulate different stakeholders, such as government ministers, interest groups such as farmer, hunter or environmental organisations, the media or the European Commission, in advanced simulation games. By allowing these actors to interact in thousands of simulations, Chapron aims to identify how and why conservation policies fail, enabling improvements before they are implemented in reality. This logic of stress-testing is borrowed from the industry where a process, for example air travel, is more likely to succeed if it repeatedly fails to fail.

The simulated stakeholders will be built using extensive data drawn from media sources, government documents, scientific publications, interest groups, opinion surveys and court rulings.

“My colleagues and I have already built a database of court decisions relating to species protection across EU Member States. These rulings reveal the arguments and positions that different actors have adopted in real-world conflicts,” says Guillaume Chapron. “We have also run a large-scale opinion survey of 10,000 respondents in the EU about nature conservation” he adds.

The simulations will also incorporate randomness in stakeholder behaviour, different personality types and unexpected events.

Wolves, bears and lynx in focus

The project will focus on case studies involving wolves, Eurasian lynx, brown bears and European bison in EU countries. All these cases are rooted in high-profile conservation issues characterised by politically contentious debates.

“Most of the case studies involve large carnivores. They provide a good basis for building the model because they are charismatic species that attract a great deal of public attention,” says Guillaume Chapron.

Guillaume Chapron has researched large carnivores for more than 20 years, often using interdisciplinary approaches. Through this new project, he will establish a research team with expertise in artificial intelligence, simulation methods, and socio-economic and political analysis.