Over the next five years, AI researcher Prof. Dr. Andreas Geiger from Tübingen, Germany, will develop new methods to support researchers in gaining new insights. For this project, he has received a Consolidator Grant from the European Research Council (ERC), one of the most prestigious European funding programs. His project, CASIDO – Computational Assistants for Scientific Discovery – will receive approximately two million euros in funding. Geiger is a professor of computer science at the University of Tübingen and a member of the Tübingen AI Center, as well as a former group leader at the Max Planck Institute for Intelligent Systems.
CASIDO: AI as a partner in the research process
The volume of scientific publications is growing rapidly. Every day, researchers have to sift through large amounts of information, classify developments, and identify connections. This overload makes it difficult to establish relevant links and develop new ideas.
This is where CASIDO comes in: The project is developing trustworthy, explainable, and human-centered AI assistants that help read, compare, and structure scientific papers, reveal gaps in research, and support creative solutions.
"We don't want AI that replaces researchers, but one that expands their capabilities," says Geiger.
The vision: tools that better organize knowledge, provide inspiration for new ideas, and thus facilitate scientific innovation.
From structured document understanding to creative ideation
CASIDO has four main areas of focus:
- Analysis of scientific texts: New, hierarchical document models are designed to better capture the structure of research articles and enable precise comparisons.
- Predicting scientific impact: By combining text content, citation networks, and visual information, models will be created that predict and explain scientific impact.
- Maps of science: Dynamic visualizations will show how research fields are developing and what trends are emerging.
- Idea generation: Creative AI assistance will highlight gaps in research and develop suggestions for new scientific approaches – always with a focus on transparency and control.
A human-scentered approach
CASIDO sees AI as a support for human creativity, intuition, and critical thinking. The focus is on collaboration between humans and machines, not on automating the research process. The developed models are made available via the Scholar Inbox platform, which is already used by many researchers today and enables realistic testing.