Final Exam

1. Why do AI systems trained on unbalanced datasets perform worse on underrepresented groups?

2. How are communities pushing back against AI perception systems?

3. What is a key consequence of higher false positive rates for one demographic group?

4. What did the 2019 NIST facial recognition study demonstrate?

5. What is necessary for effective governance of AI perception systems?

6. What is "dataset bias" and why does it matter?

7. Why is the gap between "public" and "observed" a modern privacy concern?

8. What is the fundamental difference between how AI systems perceive and how humans understand?

9. What makes traditional "consent" inadequate for ubiquitous AI perception systems?

10. How does COMPAS demonstrate how training data shapes AI bias?

11. What does the "accuracy paradox" describe?

12. What is a limitation of GDPR in regulating facial recognition systems?

13. Beyond cameras, what is the full spectrum of technologies through which modern AI systems perceive the world?

14. Why does Amazon's Sidewalk network exemplify a key concern about AI sensing?

15. What did Clearview AI's case demonstrate about data privacy regulation?