CauScien: Uncovering Causality in Science

Many scientific questions inherently possess a causal dimension, from estimating treatment effects in personalized healthcare (Greenland et al. 1999, Chakraborty et al. 2013), understanding biological mechanisms (Davey et al., 2014, Visscher et al. 2017), to evaluating the social and environmental impacts of governmental policies. At the same time, the recent surge in causal learning has unlocked the potential to extract reliable cause-and-effect relationships from complex, high-dimensional data (Schoelkopf et al., 2021). Despite promising breakthroughs in causal reasoning and discovery, translating those methods into day-to-day scientific practice remains challenging (Cadei et al., 2024), as it requires transitioning from controlled or simplified demonstrations to robust, application-driven research in complex problems.

This workshop is designed to bridge that translational gap by fostering collaboration across a broad spectrum of disciplines, including ecologists, biologists, and social scientists,  with a shared interest in both the theoretical and practical aspects of causal inference.

Overall, this workshop focuses on the following questions:

• Where does causality manifest across different applied science domains?

• When can causal learning techniques effectively accelerate scientific discovery?

• How can we best integrate causality with domain expertise and real‐world scientific data?

By promoting a bottom-up research paradigm that starts with concrete, real-world problems, we aim to (i) surface unsolved challenges that theory must address, (ii) release causal benchmark tasks that reflect real scientific constraints, and (iii) seed collaborations between domain experts and machine learning researchers that outlive the event. 

Moreover, the workshop will delve into the dual roles of causality, both as a tool for reasoning and interpretation and as a mechanism for planning and experimental design. Through discussions centered on these themes, we envision a dynamic forum for exploring the emergence of causality in applied science, identifying problem areas where causal inference and learning can drive significant improvements, and developing effective strategies for integrating causal techniques in collaboration with domain experts and applied scientific data.