Spot the Bias

End-of-lesson check

15 questions · take it digitally for instant feedback at tendril.neural-forge.io/learn/quiz/end-game-spot-the-bias-builders

1. What is the fundamental cause of bias in AI systems?

   1. The location where AI servers are stored around the world
   2. The AI's programming that deliberately favors certain groups over others
   3. The patterns and imbalances present in the data the AI was trained on
   4. The specific questions users ask when interacting with AI

2. An AI image generator consistently creates pictures of nurses as women and doctors as men. What type of bias does this demonstrate?

   1. Job bias
   2. Language bias
   3. Culture bias
   4. Image bias

3. Why do many AI language tools work better in English than in languages like Swahili or Tagalog?

   1. Much more English text was included in the AI's training data
   2. AI servers are mostly located in English-speaking countries
   3. English is a simpler language that AI can understand more easily
   4. English-speaking engineers built most AI systems

4. You want an AI image generator to create a picture of a CEO that doesn't follow the stereotypical assumption. What is the best approach?

   1. Use a different AI image generator that claims to be unbiased
   2. Specify the characteristics you want, such as 'a South Asian woman as CEO'
   3. Wait for the AI company to fix the bias in their next update
   4. Ask the AI to be fair in your prompt

5. Which statement best explains why we cannot say an AI is 'trying' to be unfair when it produces biased content?

   1. AI chooses to please the majority of its users
   2. AI has no intentions or goals—it simply reproduces patterns from its training data
   3. AI deliberately hides its biases to avoid criticism
   4. AI is aware of fairness but prioritizes accuracy over fairness

6. A company discovers their AI hiring tool keeps rejecting qualified women for engineering roles. What should they fix first?

   1. Limit which departments can use the AI tool
   2. Review and balance the historical hiring data used to train the AI
   3. Tell job applicants that the AI might be biased
   4. Add a filter to the AI's output that blocks certain words

7. Which prompt is MOST likely to produce a result that avoids stereotypical bias?

   1. Show me a software developer
   2. Show me someone who is good at math
   3. Show me a smart person at a computer
   4. Show me five software developers of different genders and ethnicities

8. A student asks an AI to 'describe a terrorist' and gets a description matching a specific ethnic group. What explains this result?

   1. The student used the word 'terrorist' which triggers automatic bias
   2. The AI is expressing its own prejudice against that group
   3. The AI's training data contained more examples connecting that group to terrorism
   4. The AI wanted to shock the student

9. What does it mean when we say AI treats 'US or UK stuff as default'?

   1. The AI refuses to answer questions about other countries
   2. The AI assumes American or British culture, norms, and examples apply everywhere
   3. The AI only works correctly when used in the US or UK
   4. The AI was built by American and British companies only

10. What is a 'stereotype' in the context of AI bias?

    1. A special type of training data used to teach AI
    2. A mathematical error in how the AI processes information
    3. An oversimplified assumption that all members of a group are the same
    4. A comparison between two different AI systems

11. Which prompt technique best helps push back against AI bias?

    1. Using simpler words so the AI understands better
    2. Asking the AI to be more fair in your request
    3. Adding specific details about the diversity you want to see
    4. Asking the AI to explain its reasoning

12. An AI chatbot gives a much longer, more detailed answer in English than in Spanish for the same question. What does this demonstrate?

    1. The AI is testing which language users prefer
    2. Language bias due to uneven training data across languages
    3. Spanish is a more difficult language for AI to process
    4. The AI prefers English-speaking users over Spanish speakers

13. A teacher asks students to critique an AI's response for potential bias. What should they look for first?

    1. Whether the answer reflects stereotypical assumptions about groups
    2. Whether the answer used proper grammar
    3. Whether the answer came from a trusted source
    4. Whether the answer is longer than expected

14. What does 'training data' mean in the context of AI?

    1. The large collection of text, images, and examples that AI learns from
    2. The software code that runs AI systems
    3. The user feedback collected after AI makes predictions
    4. The instructions users give when interacting with AI

15. Why might an AI generate an image of a man when asked to draw 'an astronaut'?

    1. The AI was programmed to prioritize male figures
    2. Astronaut is a technical term that AI interprets as male
    3. Most astronaut images in its training data showed men
    4. The AI wanted to surprise the user