Code for Towards Intersectionality in Machine Learning: Including More Identities, Handling Underrepresentation, and Performing Evaluation: Angelina Wang, Vikram V. Ramaswamy, and Olga Russakovsky (FAccT 2022)

Scripts to run

Can also add command line arguments like "--experiments 1 2" to specify which of the 4 experiments to run

• 0: Identities to include - API (Sec. 4.1)
• 1: Identities to include - Other (Sec. 4.2)
• 2: Evaluation methods (Sec. 6.1)
• 3: Structure in the data (Sec. 5.2)

python3 main.py --train_models
python3 main.py --interpret_results

Codebases drawn from

• GRY from https://github.com/algowatchpenn/GerryFair

• RDC from https://fairlearn.org/v0.5.0/api_reference/fairlearn.reductions.html

• GRP from https://github.com/frstyang/fairness-with-overlapping-groups

• RWT from https://github.com/google-research/google-research/blob/master/label_bias/README.md

• data from https://github.com/zykls/folktables

Packages required

• folktables
• sklearn
• torch
• pandas