VACA: Designing Variational Graph Autoencoders for Causal Queries

2022

Conference Paper

plg

In this paper, we introduce VACA, a novel class of vari- ational graph autoencoders for causal inference in the ab- sence of hidden confounders, when only observational data and the causal graph are available. Without making any para- metric assumptions, VACA mimics the necessary properties of a Structural Causal Model (SCM) to provide a flexible and practical framework for approximating interventions (do- operator) and abduction-action-prediction steps. As a result, and as shown by our empirical results, VACA accurately ap- proximates the interventional and counterfactual distributions on diverse SCMs. Finally, we apply VACA to evaluate coun- terfactual fairness in fair classification problems, as well as to learn fair classifiers without compromising performance.

@inproceedings{sanchez2022vaca,
  title = {VACA: Designing Variational Graph Autoencoders for Causal Queries},
  author = {Sanchez-Martin, Pablo and Rateike, Miriam and Valera, Isabel},
  booktitle = {Proceedings of the 36th AAAI Conference on Artificial Intelligence},
  volume = {7},
  pages = {8159--8168},
  publisher = {AAAI Press},
  address = {Palo Alto, CA},
  month = feb,
  year = {2022},
  doi = {10.1609/aaai.v36i7.20789},
  month_numeric = {2}
}