AI Bias and Limitations Quiz

1. An AI loan approval system denies applications at different rates for different racial groups despite using race-blind features. What explains this?

Proxy variable correlation with race

Random variation in the neural network

Insufficient training data

Activation function nonlinearity

2. How can domain experts help reduce bias in AI systems?

By identifying potential bias sources and validating fairness metrics

By increasing dataset size indefinitely

By removing all human oversight

By using only black-box models

3. What is the primary challenge in achieving fairness across multiple AI stakeholders?

Different stakeholders may have conflicting fairness definitions

Fairness requires eliminating all variability

Fairness only applies to classification tasks

Stakeholders always agree on fairness metrics

Submit

5. Which mitigation strategy addresses bias before model training?

Data preprocessing and careful feature selection

Increasing model complexity

Using ensemble methods

Tuning learning rate parameters

Submit

7. What does it mean when an AI model exhibits high accuracy but low fairness?

The model performs well overall but may discriminate against specific groups

The model is useless and should be discarded

Fairness and accuracy are impossible to balance

The test set is too small

8. Which scenario best illustrates selection bias in AI training?

Training data excludes rare but important edge cases

The model uses softmax activation

Gradient descent converges slowly

Cross-entropy loss is applied

Submit

10. What is a key limitation of AI in real-world applications?

Models cannot adapt to domain shifts or new data distributions

AI systems always improve with more data

Bias is impossible to measure

Fairness and accuracy are always perfectly aligned

11. What is the primary source of bias in most AI systems?

Biased training data reflecting historical inequities

The algorithm's mathematical complexity

Excessive computational power

The speed of the processor

Submit

13. Which of the following is a measurable fairness metric for AI systems?

Demographic parity

User satisfaction

Model interpretability

Training speed

14. What is the relationship between AI bias and fairness?

Bias always leads to unfair outcomes; eliminating bias ensures fairness

Bias and fairness are unrelated concepts

Reducing bias is necessary but insufficient for fairness; fairness requires intentional design

Fairness is only relevant in consumer applications

15. A model trained on images from wealthy neighborhoods performs poorly on urban images. This demonstrates which limitation?

Data distribution mismatch

Insufficient regularization

Batch normalization failure

Hyperparameter sensitivity

Submit

17. Which approach is most effective for detecting bias in deployed AI models?

Regular fairness audits across demographic groups

Increasing model accuracy on the test set

Using larger datasets

Simplifying the algorithm

18. What is representation bias in AI systems?

When certain groups are underrepresented in training data

When the model uses too many features

When predictions vary by time period

When loss function is improperly weighted

19. A hiring AI trained on historical recruitment data systematically rejects female applicants. This exemplifies which problem?

Underfitting

Historical bias perpetuation

Feature engineering failure

Cross-validation error

20. Which term describes when an AI model performs differently for different demographic groups?

Fairness disparity

Algorithmic bias

Data skew

Overfitting variance

AI Fundamentals Academic Recognizing AI Bias and Limitations Quiz

1. An AI loan approval system denies applications at different rates for different racial groups despite using race-blind features. What explains this?

2.

What first name or nickname would you like us to use?

2. How can domain experts help reduce bias in AI systems?

3. What is the primary challenge in achieving fairness across multiple AI stakeholders?

4. An AI system trained on historical criminal justice data perpetuates sentencing disparities. This exemplifies the concept of ____.

5. Which mitigation strategy addresses bias before model training?

6. Feedback loops in AI systems can amplify bias by ____.

7. What does it mean when an AI model exhibits high accuracy but low fairness?

8. Which scenario best illustrates selection bias in AI training?

9. Algorithmic transparency helps mitigate bias by ____.

10. What is a key limitation of AI in real-world applications?

11. What is the primary source of bias in most AI systems?

12. Proxy variables in AI models can introduce bias when they ____.

13. Which of the following is a measurable fairness metric for AI systems?

14. What is the relationship between AI bias and fairness?

15. A model trained on images from wealthy neighborhoods performs poorly on urban images. This demonstrates which limitation?

16. Confirmation bias in AI occurs when developers ____.

17. Which approach is most effective for detecting bias in deployed AI models?

18. What is representation bias in AI systems?

19. A hiring AI trained on historical recruitment data systematically rejects female applicants. This exemplifies which problem?

20. Which term describes when an AI model performs differently for different demographic groups?