This repository contains the code used to run the experiments for the paper: "Revisiting (Un)Fairness in Recourse by Minimizing Worst-Case Social Burden" (extended version is titled “Who Pays for Fairness? Rethinking Recourse under Social Burden”) accepted as an oral presentation at the The 40th Annual AAAI Conference on Artificial Intelligence 🎉
The core of this project is the implementation and evaluation of MISOB (MInimax SOcial Burden) framework for algorithmic recourse (run_misob.py). This project is built on top of the CARLA (Counterfactual And Recourse Library) library. The code has been modified and updated to be compatible with recent versions of key Python libraries (e.g., PyTorch, pandas, scikit-learn), and extended to support the MISOB framework and the experimental setup described in the paper.
We acknowledge the original CARLA authors and their foundational work in the field of recourse.