Which technique is used to explain model decisions to stakeholders?

• A. Hardware acceleration
• B. Black-box testing
• C. Data masking
• D. SHAP values or LIME

Answer: (D)

Explaining model decisions to stakeholders relies on translating a prediction into understandable contributions from the input features. SHAP values and LIME are designed for this, providing local explanations that show how each feature pushed the prediction up or down. SHAP uses Shapley values to fairly attribute the outcome to features, giving additive, consistent explanations that stay meaningful across similar cases. LIME builds a simple, interpretable surrogate model around the specific instance to approximate the complex model locally, yielding intuitive feature weights that non-technical stakeholders can grasp. These methods focus on interpretability and communicability of why a decision was made, unlike options that emphasize speed (hardware acceleration), privacy controls (data masking), or testing blind to internal reasoning (black-box testing).