We build methods and tools for causal reasoning with real-world data.
Our work focuses on pragmatic causal inference: making causal analysis usable in settings where data are messy, assumptions are imperfect, and researchers still need reliable decisions. We study front-door and proxy methods, orthogonal and debiased learning, sensitivity analysis, causal machine learning, and AI systems for practical causal reasoning.
The lab is led by Yonghan Jung at the University of Illinois Urbana-Champaign.
- Causal inference under unmeasured confounding
- Why-focused causal explanation from observational data
- What-if reasoning for interventions, policies, and decisions
- AI systems that help data scientists run causal analyses
- Debiased and orthogonal causal learners
- Front-door, proxy, and mediator-based identification
- Scalable and reproducible software for causal data science
This organization hosts selected public releases from the lab: paper PDFs, reproducibility code, posters, and project pages. Active internal development happens in private repositories before release.
Current public releases:
fdcate: public release for Debiased Front-Door Learners for Heterogeneous Effects.
Our long-term goal is to make causal analysis closer to the usability of classification and regression while preserving causal validity. When a single answer is not justified, we make uncertainty explicit instead of hiding it.