AI ethics and bias quiz

Question 1

What does 'bias' mean when we talk about an AI system?

Accepted Answer

Unfair treatment by an AI, often caused by unbalanced training data

Answer

A type of computer virus that slows down AI

Answer

A setting that makes AI run faster

Answer

The electricity an AI uses to learn

Question 2

Why is it important to use a wide variety of examples when training an AI?

Accepted Answer

Variety helps the AI work fairly for many different people and situations

Answer

Variety makes the AI use less memory

Answer

Variety makes the AI cheaper to buy

Answer

Variety stops the AI from needing electricity

Question 3

An AI learns from examples humans collect. How can human choices cause bias?

Accepted Answer

Through the data they choose, label, and leave out

Answer

Humans cannot affect AI bias at all

Answer

Only by writing slow code

Answer

Only by turning the computer off

Question 4

What is 'fairness' in the context of AI decisions?

Accepted Answer

Treating people without unfair favoritism or harm

Answer

Making the AI answer as fast as possible

Answer

Letting the AI choose its own rules

Answer

Charging everyone the same price for software

Question 5

If a hiring AI was trained mostly on resumes of men, what problem might appear?

Accepted Answer

It might unfairly favor men and rank women lower

Answer

It would refuse to read any resumes

Answer

It would only work on weekends

Answer

It would automatically hire everyone

Question 6

What does 'representative data' mean and why does it matter for fairness?

Accepted Answer

Data that includes all relevant groups and situations fairly

Answer

Data stored in a representative's office

Answer

Data that is only about famous people

Answer

Data that has been deleted

Question 7

What is 'privacy' and why should AI builders care about it?

Accepted Answer

Privacy is a person's right to control their personal information

Answer

Privacy is how fast an AI runs

Answer

Privacy is the brand name of an AI

Answer

Privacy means the AI never makes mistakes

Question 8

Why might collecting people's data without their permission be unethical?

Accepted Answer

It ignores people's right to control their own information

Answer

It uses too much storage

Answer

It makes the AI slower

Answer

It is always perfectly fine

Question 9

What is 'transparency' in AI, and why is it valuable?

Accepted Answer

Being able to understand how and why an AI made a decision

Answer

Making the AI screen see-through

Answer

Hiding the AI's code from everyone

Answer

Making the AI invisible to users

Question 10

What does it mean for an AI system to be 'accountable'?

Accepted Answer

Humans are responsible for what the AI does and for fixing harm

Answer

The AI counts numbers very fast

Answer

The AI never needs updates

Answer

The AI pays its own electricity bill

Question 11

A photo AI labels a picture of a nurse as 'woman' and a picture of a doctor as 'man'. What is this an example of?

Accepted Answer

Stereotype bias learned from data

Answer

A hardware failure

Answer

A privacy violation

Answer

Faster processing

Question 12

What is a 'feedback loop' that can make AI bias worse over time?

Accepted Answer

A cycle where biased AI decisions create biased new data, worsening bias

Answer

A cable that connects two computers

Answer

A way to make AI run backwards

Answer

A sound the computer makes when overheating

Question 13

Why can an AI be very accurate overall but still be unfair to a small group?

Accepted Answer

Because high average accuracy can hide poor results for small groups

Answer

Because small groups don't use computers

Answer

Because AI cannot be accurate

Answer

Because accuracy and fairness are the same thing

Question 14

What is a simple way to check if an AI is fair across different groups?

Accepted Answer

Measure and compare accuracy separately for each group

Answer

Only check the total accuracy

Answer

Ask the AI if it is fair

Answer

Run the AI faster

Question 15

What is 'misinformation' and how can AI accidentally spread it?

Accepted Answer

False information that can be spread by confident but wrong AI outputs

Answer

A way to store data safely

Answer

A type of fair AI

Answer

A faster internet connection

Question 16

What is a 'deepfake' and why does it raise ethical concerns?

Accepted Answer

AI-made fake media that can deceive people and harm reputations

Answer

A very deep database

Answer

A type of strong password

Answer

A computer that is buried underground

Question 17

Why should humans review important decisions made by AI, like medical or legal ones?

Accepted Answer

Because AI can make biased mistakes and important decisions affect people's lives

Answer

Because AI is always wrong

Answer

Because humans type faster

Answer

Because AI gets tired

Question 18

What does 'garbage in, garbage out' mean for AI fairness?

Accepted Answer

Bad or biased input data leads to bad or biased AI results

Answer

AI should be kept near a trash can

Answer

AI produces actual garbage

Answer

Garbage data makes AI faster

Question 19

How can leaving a group out of the training data harm that group?

Accepted Answer

The AI may make more mistakes for that unseen group

Answer

It helps that group get better service

Answer

It makes the AI run faster for everyone

Answer

It has no effect at all

Question 20

What is 'consent' and why does it matter when using someone's photos to train AI?

Accepted Answer

A person's informed agreement to let their information be used

Answer

A setting that brightens photos

Answer

A type of camera lens

Answer

The price paid for a photo

Question 21

If an AI translation tool always translates 'engineer' as male, what kind of bias is shown?

Accepted Answer

Gender bias

Answer

Hardware bias

Answer

Speed bias

Answer

Color bias

Question 22

Why is it a problem if we cannot explain how an AI made a decision?

Accepted Answer

We cannot check if it was fair or correct, or fix mistakes

Answer

It makes the AI run slower

Answer

It uses more electricity

Answer

It is never a problem

Question 23

What is the difference between an AI being 'accurate' and being 'fair'?

Accepted Answer

Accuracy is about correct answers overall; fairness is about just treatment of groups

Answer

They mean exactly the same thing

Answer

Fairness is about speed; accuracy is about cost

Answer

Both only matter for games

Question 24

How can an AI recommendation system create a 'filter bubble'?

Accepted Answer

Showing only similar content so you miss other viewpoints

Answer

A bubble that cleans the screen

Answer

A way to block all ads

Answer

A faster way to load videos

Question 25

Why should AI builders test their system before letting the public use it?

Accepted Answer

To find mistakes and unfair results before they harm real people

Answer

To make the AI more expensive

Answer

Because the law forbids using untested AI for games

Answer

To slow the AI down

Question 26

What is 'data labeling' and how can it introduce bias?

Accepted Answer

Tagging examples for the AI to learn from, which can carry human bias

Answer

Sticking price tags on computers

Answer

Naming files randomly

Answer

Deleting old data

Question 27

Imagine a school uses AI to predict which students will succeed. What ethical risks should they consider?

Accepted Answer

Unfair labeling from biased data, privacy risks, and over-reliance on the AI

Answer

Only the cost of the computers

Answer

That the AI might be too fast

Answer

There are no ethical risks

Question 28

What does 'inclusive design' mean in AI?

Accepted Answer

Building AI to work fairly for many kinds of people, including those often left out

Answer

Designing AI only for experts

Answer

Making the AI look colorful

Answer

Designing AI to run on one computer only

Question 29

Why is it risky to fully trust an AI just because it sounds confident?

Accepted Answer

AI can sound confident even when it is wrong

Answer

Confidence always means the answer is right

Answer

Confident AI uses less power

Answer

Confidence is illegal in AI

Question 30

How can using diverse teams of people help reduce bias in AI?

Accepted Answer

Different viewpoints help spot unfairness and blind spots others miss

Answer

Diverse teams make AI run faster

Answer

Diverse teams need fewer computers

Answer

It has no effect on bias

Test yourself