2. Fairness Extensions in Optimal Policy

The OptimalPolicy class in the mcf includes experimental features for fairness adjustments, accessible through the fairscores method. These features are designed to ensure that policy scores are fair with respect to certain protected variables. The fairness adjustments are based on the upcoming work by Bearth, Lechner, Mareckova, and Muny (2024).

This method can be configured using several parameters to control the type and extent of fairness adjustments.

2.1. Example

To use the fairness adjustments, configure the OptimalPolicy class with the appropriate parameters and call the fairscores method on your data. This will return a DataFrame with adjusted policy scores that account for fairness considerations.

import os
from mcf.example_data_functions import example_data
from mcf.optpolicy_functions import OptimalPolicy

# Generate data
training_df, prediction_df, name_dict = example_data(
    obs_y_d_x_iate=1000,
    obs_x_iate=1000,
    no_features=5,
    no_treatments=2,
    seed=12345,
    type_of_heterogeneity='WagerAthey',
    descr_stats=True
)

# Define parameters for OptimalPolicy
params = {
    'var_d_name': 'treat',
    'var_polscore_name': ('y_pot0', 'y_pot1'),
    'var_x_name_ord': ('x_ord0', 'x_ord1'),
    'var_protected_name_ord': ('x_ord0'),  # Specify at least one protected feature
    'gen_outfiletext': "OptPolicy_Simple_Example",
    'gen_outpath': os.getcwd() + '/outputOPT'
}

# Initialize and adjust fairness scores
myoptp_fair = OptimalPolicy(**params)
training_fair_df, fairscore_names, _, _ = myoptp_fair.fairscores(training_df.copy(), data_title='training')

# Update params to use the fair scores
params['var_polscore_name'] = tuple(fairscore_names)

# Solve for optimal allocation rule
alloc_train_df, _, _ = myoptp_fair.solve(training_fair_df.copy(), data_title='training fair')

# Evaluate the allocation
results_eva_train, _ = myoptp_fair.evaluate(alloc_train_df, training_fair_df.copy(), data_title='training fair')

2.1.1. Note

These features are experimental and may require further testing and validation. For more details or additional parameters, please consult the API documentation.