Counterfactual Situation Testing

This is the repository for Counterfactual Situation Testing: From Single to Multidimensional Discrimination, published in the Journal of Artificial Intelligence (JAIR). The datasets are located in the data/ folder. The scripts get_data_< > prepare each dataset. All other scripts are in the src/ folder. The scripts get_cf_data_< > generate the counterfactual dataset. In the case for the law school data, a second get_cf_data_ script is used for the intersectional discrimination case. For this dataset, in particular, we generate the counterfactual dataset in R also to be able to compare the 'frequentist' and 'Bayesian' approaches of the abduction step. The latter approach uses RStan. The scripts run_exp_< > run the experiments. Finally, the scripts analysis_< > contain additional analysis. The results, including relevant figures, are in the results/ folder.

EAAMO'23 version

An earlier version of this work, Counterfactual Situation Testing: Uncovering Discrimination under Fairness given the Difference, was published at ACM EAAMO 2023. See the eeamo2023 branch for details. The code has changed for the journal version; please use the master branch for the latest code.

References

Counterfactual Situation Testing: From Single to Multidimensional Discrimination. Jose M. Alvarez, and Salvatore Ruggieri. Journal of Artificial Intelligence Research (JAIR), 82, 2279-2323, 2025.

If you make use of the code, the CST algorithm, or the generated synthetic data in your work, please cite the following paper:

@article{DBLP:journals/jair/AlvarezR25,
  author       = {Jos{\'{e}} M. {\'{A}}lvarez and
                  Salvatore Ruggieri},
  title        = {Counterfactual Situation Testing: From Single to Multidimensional
                  Discrimination},
  journal      = {J. Artif. Intell. Res.},
  volume       = {82},
  pages        = {2279--2323},
  year         = {2025}
}