Explainable AI

What I do

• Apply XAI techniques to models
• Generate model-agnostic explanations
• Create interpretable features
• Build trust in AI systems
• Handle fairness and bias

When to use me

Use me when:

• Regulatory compliance (GDPR, EU AI Act)
• High-stakes decisions
• Debugging model behavior
• Building trust with users

Key Concepts

XAI Methods

• SHAP: Shapley additive explanations
• LIME: Local interpretable model-agnostic
• Counterfactuals: What-if analysis
• Feature importance: Permutation, tree-based

SHAP Framework

import shap

# Kernel SHAP for any model
explainer = shap.KernelExplainer(model.predict, X_background)
shap_values = explainer.shap_values(X_test)

# Tree SHAP for tree-based
explainer = shap.TreeExplainer(model)
shap_values = explainer.shap_values(X_test)

# Visualization
shap.force_plot(
    explainer.expected_value, 
    shap_values[0,:], 
    X_test.iloc[0,:]
)

Counterfactual Explanations

from dice_ml import Explainer

# Create explainer
explainer = Explainer(model, X_train, outcome_class="target")

# Generate counterfactuals
explanation = explanation.generate_counterfactuals(
    X_test[0:1], 
    total_CFs=3, 
    desired_class="opposite"
)

Fairness

• Demographic parity
• Equalized odds
• Individual fairness
• Bias detection metrics