Partial Dependence Plots

Partial Dependence Mutates the Dataset

Training a Global Surrogate

Limitations of Explanation Methods

Stakeholders and Audiences for Explanations

Interpretability Tools and Libraries

Evaluating the Quality of Explanations

Attention as an Explanation and Its Caveats

Surrogate Models for Explaining Black Boxes

Counterfactual Explanations

Saliency Maps for Deep Models

Individual Conditional Expectation Plots

Partial Dependence Plots (PDP)

LIME: Local Model Explanations

SHAP Values and Shapley Explanations

Permutation Feature Importance

Feature Importance Methods

The Accuracy-Interpretability Tradeoff

White-Box Models: Linear & Trees

Model-Agnostic vs Model-Specific Methods

Global vs Local Explanations

Intrinsic vs Post-Hoc Interpretability

Why Model Interpretability Matters

Choosing an Attribution Baseline

Anchor Rule Precision

Normalizing Attention Weights

Faithfulness by Feature Removal

Correlation as Attribution

Counterfactual Search Loop

Stakeholder-Appropriate Explanations

Example-Based Explanations

SHAP and Shapley Values

Accuracy-Interpretability Trade-off

Surrogate Models

Interpreting Decision Trees

Right to Explanation

Faithfulness of Explanations

Partial Dependence Plots

ICE Plots

Attention as Explanation

Interpretability vs Explainability

Counterfactual Explanations

Saliency Maps

Global vs Local Explanations

What Interpretability Means

Model-Agnostic vs Model-Specific

Interpretable vs Black-Box Models

Permutation Feature Importance

Feature Importance Concept

LIME Explanations

Linear Model Coefficients

ICE Curves Collapsed Into a PDP

Grad-CAM Averages Over Wrong Axis

Integrated Gradients Missing Input Difference

LIME Surrogate Ignores Proximity Weights

Perturbation Explanation Predicts Wrong Input

Ranking SHAP Contributions by Impact

Saliency Map Keeps Signed Gradients

Odds Ratio from Logistic Coefficient

Permutation Feature Importance Sign

SHAP Additivity Check

Global Importance Over Whole Dataset

Averaging Shapley Marginal Contributions

Reconstructing a Prediction from SHAP

Classifying Intrinsic vs Post-Hoc Methods

Most Important Linear Coefficient

Linear Model Coefficients

Evaluating Explanation Quality

Grad-CAM for CNNs

Surrogate Models

Chain-of-Thought as Explanation

ICE Plots

Faithfulness vs Plausibility

Pitfalls of Explanation Methods

Attention as Explanation

Saliency Maps

Counterfactual Explanations

White-Box Models

Why Interpretability Matters

Intrinsic vs Post-hoc

LIME

Feature Importance

Permutation Feature Importance

Interpretability vs Explainability

SHAP

Global vs Local Explanations

Accuracy-Interpretability Tradeoff

Model-Specific vs Model-Agnostic