8.4. Fairness Simulation Study 1

This example demonstrates the use of PiML for fairness testing based on the integrated solas-ai APIs. The SolasSimu1 data is modified from the Friedman demo data. The covariates used for modeling are x1, x2, …, x5, the response Label is binary and it is a classification problem. The rest variables are demographic variables used for testing fairness. The data and APIs are contributed by Solas-AI.

Click the ipynb links to run examples in Google Colab.

8.4.1. Load and Prepare Data

from piml import Experiment
exp = Experiment()

# Choose SolasSimu1
exp.data_loader()

# Exclude demographic features one-by-one: "Minority", "Majority", "Age >= 62", "Age < 62"
# Excluded features will show in grey color in the table
exp.data_summary()

# Prepare dataset with default settings
exp.data_prepare()

# Exploratory Data Analysis
exp.eda()

8.4.2. Train ML Model(s)

# Choose GAM with default settings, click run;
# When training is finished, register the model
exp.model_train()

8.4.3. Fairness Testing

Suggested Procudure:

1. First select a registered model (in this case, GAM)

2. Group Setting:

   • Set Add Category=”By weights”, Enter “Race” in Category Name; Select “Majority” as reference; Select “Minority” as protected; Click on Add Button

   • Set Add Category=”By weights”, Enter “Age” in Category Name; Select “Age<62” as reference; Select “Age>=62” as protected; Click on Add Button

3. Switch to other tabs to view metrics and segmented breakdowns.

exp.model_fairness_solas()