The ever-increasing use of data-driven models (whether classical machine learning or novel deep learning models, such as LLMs) to automate software engineering (SE) is accentuating the need to distinguish between causal effects and spurious correlations. While the application of causal methods in SE is gradually attracting more attention, it is not yet widely used. This 90 min. tutorial provides an introduction to causal inference techniques for observational data based on causal graphs and its application to SE. The goal is to enable SE researchers to apply such methods to their own use cases and streamline the knowledge and use of causal inference within the SE community. Causal inference methods focus on the question of identifying and estimating the causal impact of a variable of interest on another one even when the data was not collected in a randomized controlled experiment (i.e., using observational data). Besides the theory, this tutorial will also showcase how such methods have been applied in SE and it will give pointers to libraries in Python and R.

@inproceedings{SiebertFHP26,
  title = {Introduction to Causal Inference for Software Engineering},
  author = {Julien Siebert and Julian Frattini and Hans-Martin Heyn and Roberto Pietrantuono},
  booktitle = {FSE},
  year = {2026},
  doi = {10.1145/3803437.3804898}
}