Treatment Equality

Overview

Calculates difference of FN / FP betweeen priv and unpriv groups

Formula

Treatment Equality = False Negatives / False Positives

Source

Usage

Manually

Python

# Calculate treatment equality
result = treatment_equality(df, protected_attribute, privileged_group, labels, positive_label, y_true)

print("Treatment Equality:", result)

Using Fairness Object

result = (fo.compute(treatment_equality))

Results

Treatment Equality: -0.9999999999666667

These results are obtained by using the input data given in the Create Example Data page under Getting Started

Interpretation

The primary concept behind this metric is to evaluate if errors have a disproportionate impact on one group compared to another, even when accuracy is equal between the groups.