Proxy Variable Detector
Identify features that serve as proxies for protected attributes in your dataset.
Features
Features in the dataset, colored by proxy risk level for the selected protected attribute.
Proxy Correlation Map
Variance Inflation Factor (VIF) & Multicollinearity
VIF detects features that are redundant with each other — high VIF means the feature's information overlaps with other features, amplifying proxy effects.
Proxy Alerts
Recommendations
Understanding Proxy Detection
Proxy variables are features correlated with protected attributes. Simple Pearson correlation only captures linear relationships. This tool also shows Spearman rank correlation (captures monotonic relationships) and VIF (detects multicollinearity amplifying proxy effects). A feature with both high proxy correlation and high VIF is especially dangerous because its information overlaps with other discriminatory features.