This directory does not ship the dataset. The Law School data used by the experiments is third-party data with its own usage terms, so it is fetched on demand rather than redistributed here.

# from the repository root
LAWSCHOOL_DATA_URL="https://<source-you-are-entitled-to-use>/law_school.csv" \
    python data/get_lawschool_data.py

This writes data/LawSchool.csv (pipe-separated, sep='|') with the columns:

race | sex | LSAT | UGPA | ZFYA | race_nonwhite | race_simpler

Then regenerate the counterfactual datasets (they are derived, not downloaded):

Rscript src/get_cf_data_law_school.R
# produces: cf_LawSchool_lev3_doMale.csv, cf_LawSchool_lev3_doWhite.csv

The Law School dataset records, for ~21,790 US law-school students, academic metrics (LSAT, UGPA, ZFYA) together with coarse demographic attributes (race, sex). It contains no names or direct identifiers; it is an anonymized, aggregated research dataset originally collected in the United States in the 1990s. The demographic columns are special-category attributes under GDPR Art. 9, but the dataset is anonymized and out of the scope of GDPR for these records (Recital 26). Confirm anonymization and your redistribution rights before publishing any copy.

Original data:

Wightman, L. F. (1998). LSAC National Longitudinal Bar Passage Study. LSAC Research Report Series. Law School Admission Council. Distributed by SEAPHE: http://www.seaphe.org/databases.php

Pre-processed variant (adds ZFYA and the simplified race columns), as used in the algorithmic-fairness literature:

Kusner, M. J., Loftus, J. R., Russell, C., & Silva, R. (2017). Counterfactual Fairness. Advances in Neural Information Processing Systems (NeurIPS) 30.

If you use this repository, please also cite the accompanying paper — see CITATION.cff at the repository root.