Justicia

This is the implementation of our AAAI 2021 and 2022 papers where we have proposed a formal approach to verify the fairness of machine learning classifiers.

Documentation

Python tutorials are available in doc.

Install

• Install python dependencies (prerequisite) pip install -r requirements.txt
• Install the python library pip install justicia

Other dependencies

• PGMPY

• SSAT solver. Checkout to the compatible version.

  git clone https://github.com/NTU-ALComLab/ssatABC.git
  cd ssatABC
  git checkout 91a93a57c08812e3fe24aabd71219b744d2355ad
  

• PySAT

Citations

Please cite following papers.

@inproceedings{ghosh2022algorithmic,
author={Ghosh, Bishwamittra and Basu, Debabrota and Meel, Kuldeep S.},
title={Algorithmic Fairness Verification with Graphical Models},
booktitle={Proceedings of AAAI},
month={2},
year={2022},
}

@inproceedings{ghosh2021justicia,
author={Ghosh, Bishwamittra and Basu, Debabrota and Meel, Kuldeep S.},
title={Justicia: A Stochastic {SAT} Approach to Formally Verify Fairness},
booktitle={Proceedings of AAAI},
month={2},
year={2021},
}

Contact

Bishwamittra Ghosh (bghosh@u.nus.edu)

Issues, questions, bugs, etc.

Please click on "issues" at the top and create a new issue. All issues are responded to promptly.