Module Check: Bias and Fairness in AI
You have completed nine lessons on bias and fairness in AI. Before you move on, this lesson pulls all the key ideas together. Start with the flashcard review to warm up your memory, then work through the quizzes to test your understanding across every lesson in the module. End with the synthesis activity, which asks you to step back and see the full picture.
Flashcard Review — Key Terms
Flashcards — click each card to reveal the answer
Module Quiz
What distinguishes bias from a random mistake?
An AI trained on loan approval records from the 1990s shows worse outcomes for female applicants today. What is the most likely cause?
A voice assistant accurately recognizes American English at 97 percent accuracy but only 71 percent accuracy for Nigerian English. What best explains this gap?
A hiring AI does not include 'gender' as a feature. However, it includes 'college sports team captain' and 'fraternity or sorority membership.' Why might this still produce gender-biased results?
A criminal justice AI is 90 percent accurate for Group A and 90 percent accurate for Group B overall. But among people who were actually low risk, it correctly identifies 95 percent of Group A as low risk but only 72 percent of Group B as low risk. Which fairness metric is violated?
Which statement best captures the fairness impossibility result?
Synthesis Activity
The Full Picture: A Letter to a Future AI Developer
- You are writing a letter to a student one year younger than you who is about to start studying AI. They have heard that AI is biased and they want to understand what that really means. Your letter must cover all of the following, in your own words:
- 1. What bias is and why it is more than just a mistake. Use one specific example from this module.
- 2. The two main ways bias enters AI systems, with a one-sentence explanation of each.
- 3. One documented real-world case of AI bias and why it mattered to the people it affected.
- 4. What a representation gap is and why it means the people most harmed are often the people least heard.
- 5. The fairness impossibility problem — explain it so clearly that someone who has never taken this course can understand it.
- 6. One thing engineers can do, one thing organizations can do, and one thing users can do to make AI fairer.
- 7. Your honest opinion: do you think AI systems will eventually be fair enough to use in high-stakes decisions like criminal sentencing or medical diagnosis? Explain your reasoning.
- Your letter should be written as if to a real person — engaged, clear, and in your own voice. Minimum five paragraphs.