Core Protocol

Explainability (SHAP)

Understand exactly why your model makes biased predictions. We utilize SHAP (SHapley Additive exPlanations) to break down the impact of every feature on the final prediction.

Global Feature Importance

Calculates the average magnitude of the SHAP values across all samples. This reveals which features are generally the most important to the model's decision-making process overall. If a sensitive attribute (or a known proxy) appears at the top of this list, the model is heavily relying on demographic data.

Beeswarm Plots

Provides a high-density visualization of how the value of each feature impacts the model output. Each dot represents a single prediction. Color indicates the feature value (e.g., age: high or low), and the horizontal axis shows the SHAP value (impact on prediction).

Example: If older age (red dots) consistently pushes the SHAP value negative for a loan approval model, the model is systematically biased against older applicants.