NEW
Font size
WorksheetsIntroduction to XAI
Total questions: 18
Worksheet time: 9mins
What is the primary goal of Explainable Artificial Intelligence (XAI)?
To increase model accuracy only
To reduce computational cost
To make AI decisions understandable to humans
To replace traditional machine learning
Why is explainability important in AI systems?
To improve hardware performance
To build trust, accountability, and transparency
To eliminate bias completely
To automate decision-making
Which of the following best describes a “black-box” model?
A model with low accuracy
A model that cannot be deployed
A model whose internal logic is difficult to interpret
A model that uses open-source data
Which model is generally considered inherently interpretable?
Deep Neural Network
Random Forest
Convolutional Neural Network
Linear Regression
Explainable AI techniques are mainly used to explain:
Only interpretable models
Only rule-based systems
Complex or black-box models
Statistical tests
Which statement is TRUE about interpretability and explainability?
They are exactly the same concept
Interpretability focuses on model structure, explainability on outputs
Explainability applies only to neural networks
Interpretability always requires post-hoc methods
Feature importance helps users understand:
How fast the model runs
Dataset size
Which features contribute most to predictions
Model fairness
SHAP values are based on which concept?
Bayesian probability
Game theory (Shapley values)
Gradient descent
Decision trees
LIME explains model predictions by:
Training a new global model
Using gradients
Approximating the model locally with a simple model
Visualizing neural layers
Which statement best differentiates SHAP from LIME?
SHAP provides consistent global and local explanations
SHAP is faster but less accurate
SHAP is faster but less accurate
LIME is model-specific, SHAP is not
A local explanation focuses on:
Overall model performance
A single prediction
Dataset distribution
Model training time
Which XAI method can provide both global and local explanations?
Confusion matrix
Accuracy score
K-means
SHAP
Global explanations are useful for:
Debugging individual predictions
Labeling data
Improving hardware optimization
Understanding general model behavior
XAI is especially important in high-stakes domains such as:
Music recommendation
Video streaming
Online gaming
Finance and healthcare
Which of the following is NOT a benefit of XAI?
Increased user trust
Better regulatory compliance
Guaranteed fairness
Improved transparency
How does XAI support ethical AI?
By improving GPU speed
By explaining and auditing AI decisions
By removing all bias
By automating ethics
Which statement best reflects the role of humans in XAI?
Humans should not question AI decisions
Humans only collect data
Humans replace AI models
Humans validate and interpret AI explanations
Explainability in AI should be:
Optional
Used only in research
Integrated into real-world AI systems
Avoided due to complexity
