Adapting Static Fairness to Sequential Decision-Making: Bias Mitigation Strategies towards Equal Long-term Benefit Rate

ICML 2024

Yuancheng Xu* · Chenghao Deng* · Yanchao Sun · Ruijie Zheng · Xiyao Wang · Jieyu Zhao · Furong Huang

[Paper] · [Project page] · [X (Twitter)]

The codebase for the paper "Equal Long-term Benefit Rate: Adapting Static Fairness Notions to Sequential Decision Making" (https://arxiv.org/abs/2309.03426) by Yuancheng Xu*, Chenghao Deng*, Yanchao Sun, Ruijie Zheng, Xiyao Wang, Jieyu Zhao, Furong Huang.

The implementation of ELBERT Policy Optimization (ELBERT-PO) is provided.

Environment set-up

First, install Anaconda to set up virtual environment. Then, run:

conda env create -f elbert.yaml
conda activate elbert
pip install -r requirements.txt

Running ELBERT and other baselines

The .scripts/ folder includes bash scripts for ELBERT and other baselines (G-PPO, R-PPO, A-PPO) in five enviroments:

Lending

• ELBERT

bash scripts/lending_elbert.sh

• Baseline (G-PPO, R-PPO, A-PPO)

bash scripts/lending_original.sh

Infectious control, orginal version

• ELBERT

bash scripts/infectious_original_env_elbert.sh

• Baseline (G-PPO, R-PPO, A-PPO)

bash scripts/infectious_original_env_original.sh

Infectious control, harder version

• ELBERT

bash scripts/infectious_harder_env_elbert.sh

• Baseline (G-PPO, R-PPO, A-PPO)

bash scripts/infectious_harder_env_original.sh

Attention allocation, orginal version

• ELBERT

bash scripts/attention_original_env_elbert.sh

• Baseline (G-PPO, R-PPO, A-PPO)

bash scripts/attention_original_env_original.sh

Attention allocation, harder version

• ELBERT

bash scripts/attention_harder_env_elbert.sh

• Baseline (G-PPO, R-PPO, A-PPO)

bash scripts/attention_harder_env_original.sh

Comments

Our codebase is based on the following repo:

• Policy Optimization with Constraint Advantage Regularization
• Google ML-fairness-gym

Thanks for their efforts to make the code public!

Citing

Please cite our work if you find it is helpful:

@inproceedings{xuadapting,
  title={Adapting Static Fairness to Sequential Decision-Making: Bias Mitigation Strategies towards Equal Long-term Benefit Rate},
  author={Xu, Yuancheng and Deng, Chenghao and Sun, Yanchao and Zheng, Ruijie and Wang, Xiyao and Zhao, Jieyu and Huang, Furong},
  booktitle={Forty-first International Conference on Machine Learning}
}