This repository is the official implementation of Auditing Fairness under Unobserved Confounding (AISTATS 2024), containing code to implement our bounds and perform semi-synthetic data experiments.

To reproduce our results, the following command can be simply used:

python semi_synth.py

To install requirements, setup a conda environment using the following commands:

conda create -n fairness python=3.11 pip 
conda activate fairness
pip install -r requirements.txt

This repository is licensed under the terms of the MIT License.

For more details, refer to the accompanying paper: Auditing Fairness under Unobserved Confounding. If you have questions, please feel free to reach us at yewonb@cs.cmu.edu or to open an issue.