Fairness Checkers
You have learned so much about fairness and AI in this module. You know what fairness means. You know that AI learns from examples. You know that unfair examples can create unfair AI. You know that everyone should be included. You know how to speak up. Now it is time to put all of that together. In this lesson, you are a fairness checker — someone whose job is to look carefully at examples and decisions and ask: is this fair? Who might be left out? What could be better? Fairness checkers are real people. Some of them are called AI ethics researchers. Some of them are called fairness engineers. And some of them are kids just like you who notice things and ask questions. Let us get to work.
The Fairness Checker's Toolkit
Every fairness checker needs a set of questions to ask. Here are the most important ones. Question 1: Who is included? Look at the examples or decisions. What kinds of people are represented? Are there many different ages, backgrounds, languages, and abilities shown — or mostly just one type? Question 2: Who is missing? This is the hardest question because you have to notice what is not there. If all the examples look the same, ask: who would look different? Where are they? Question 3: Who might be harmed? If the AI makes a mistake, who suffers? Is the harm spread evenly, or does it fall harder on some groups than others? Question 4: Whose perspective is not heard? Sometimes an AI is built without asking the communities it will affect. If the people being helped were not part of the design, important things can get missed. Question 5: Can this be fixed? If there is a fairness problem, what would fixing it look like? Who would need to be involved?
Fairness checkers ask five key questions: Who is included? Who is missing? Who might be harmed? Whose perspective is not heard? And how can this be fixed? Anyone can be a fairness checker — including you.
Fairness Checker Challenge
- You are now an official fairness checker. Work through the three scenarios below. For each one, answer the five fairness checker questions.
- SCENARIO 1: Photo Faces
- An AI is being trained to recognize faces for a school attendance system. The researchers collect 10,000 photos. Later, someone looks at the photos and notices: 9,500 of them are of adults. Only 500 are of children. Almost all of the children in the photos have similar skin tones and hair types.
- Your task: Who is included? Who is missing? Who might be harmed by mistakes? Whose perspective was not heard? How would you fix the example set?
- SCENARIO 2: Story Suggestions
- A school library AI suggests books for students to read based on what other students liked. A student named Fatima notices the AI never suggests books with characters who share her cultural background. When she looks at the suggestion data, she sees most of the example readers were from one part of the country.
- Your task: Who is included? Who is missing? Who might feel left out? Whose perspective was not heard? What would make this fairer?
- SCENARIO 3: Weather Warning AI
- A city builds an AI to send emergency weather alerts. The AI sends alerts as text messages in English only. The city has many residents who speak Spanish, Somali, and Mandarin. The AI also only sends text alerts — no audio alerts for people with visual impairments.
- Your task: Who is included? Who might not get the warning in time? Who could be harmed by the gap? What two changes would you make first?
- Write or draw your answers for at least TWO of the three scenarios. Then share your thinking with a trusted adult. Ask them: are there fairness problems you see that I might have missed?
Being a fairness checker is not about criticizing everyone who builds AI. Most people who build AI are trying their best. But trying your best is not always enough — sometimes problems hide in places nobody thought to look. Fairness checkers are helpers. They are partners to the people building AI. They bring fresh eyes and ask the questions that someone deep in the work might miss. When you practice asking fairness questions today, you are practicing a skill that the world genuinely needs more of.
Match each fairness checker question to what it helps you find.
Terms
Definitions
Drag terms onto their definitions, or click a term then click a definition to match.
Nobody spots every fairness problem on the first look. The more you practice asking these questions, the better your eyes and brain get at noticing what is there — and what is missing. Keep practicing!
A fairness checker is reviewing an AI trained to recognize children's faces but finds that 95% of the training photos are of adults. Which fairness checker question is most directly answered by this finding?
What is the job of a fairness checker?