Counterfactual Fairness with Partially Known Causal Graph

This is the reproduction of the paper https://arxiv.org/abs/2205.13972.

To reproduce results on synthetic datasets, run the file run.sh in the working directory.

Examples and more explanations on commands in run.sh:

• python data_generator.py $j $((2*j)) to generate the observational and counterfactual data for the setting (j)Node(2*j)Edge;
• Rscript main.R $j $((2*j)) $i to fit the models using the data generated by the i_th causal graph in the setting (j)Node(2*j)Edge;
• Rscript Simu_results_filter_and_processor.R to obtain Table 1. Numeric results for RMSE and Unfairness.
• Rscript Simu_filtered_Boxplot.R to obtain Figure 1. Boxplot for RMSE and Unfairness.
• The resulting tables and boxplots can be found in the folder 'Repository'.

To reproduce results on real datasets, run Rscript main_RealData_RMSE_and_Unfairness.R to obtain Table 2; run Rscript main_RealData_Density_plots.R to obtain Figure 3.