Fair or Unfair?

What You Will Learn

• What fairness means when AI makes decisions
• How unfair data can lead to unfair AI
• How to start spotting when AI might be treating people differently

What Fairness Means in AI

Fairness means people are treated in a way that is just and equal.
In real life, people argue about what “fair” means, but most agree that things like race, gender, or family background should not decide who gets opportunities.

When AI is used to help make decisions, fairness still matters.
If an AI helps decide who gets into a program, who gets shown certain ads, or who is flagged for extra checks, unfair patterns can hurt real people.

AI does not have a sense of justice.
It does not “want” to be fair or unfair.
It simply copies patterns in the data it is trained on.

How Unfair Patterns Get Into AI

AI learns from data taken from the real world.
The real world is not always fair.

If, for many years, a group of people had fewer chances to go to good schools or get certain jobs, the data might show those people with lower incomes or fewer “success stories,” even if they had the same talent or effort.

If this data is used to train an AI that predicts who will be “successful,” the AI may quietly learn:

• People like this got more chances in the past
• People like that got fewer chances in the past

Then the AI may recommend repeating the same pattern.

No one told the AI “be unfair.”
It simply followed the pattern in the data.

Examples of Fair and Unfair AI Use

AI can be used fairly.

• Helping all students find learning materials at their level
• Translating content so more people can understand
• Detecting eye problems from images for people in many countries

AI can be used unfairly or in a risky way.

• A hiring system trained mostly on past hires from one group
• A face recognition system that works better on some skin tones than others
• A school system that gives fewer chances to students from certain neighborhoods

If an AI makes more mistakes for one group of people, that is unfair, even if the average accuracy looks high.

Who Decides What Is Fair?

There is no single formula for fairness that everyone agrees on.
Different people care about different aspects.

Some might focus on:

• Everyone having the same chance to be selected
• The system making the same error rates for every group
• Making extra effort to help groups who were treated unfairly in the past

Because fairness is complicated, people must talk about it and set rules for how AI is allowed to be used.

This includes:

• Checking AI results separately for different groups
• Asking whether historical data repeats discrimination
• Involving many voices in the design of AI, not only one small group

How You Can Start Noticing Fairness Issues

Even as a young person, you can begin to notice signs that something might be unfair.

You can ask:

• Who seems to benefit from this AI system?
• Is there a group that always seems to get worse results?
• Does the system give people a way to ask questions or appeal decisions?

You do not need to solve everything yourself.
Simply noticing and asking questions is a strong first step.

Summary Points

• AI does not understand fairness; it copies patterns in the data it sees
• If past data is unfair, AI can repeat or worsen that unfairness
• Fairness can mean different things, but treating groups very differently is a warning sign
• People must test AI on different groups, set rules, and stay alert for unfair outcomes
• You can begin to spot fairness issues by asking who benefits and who is left out

Practical Tasks

• Think of one situation where you would be upset if a human treated you unfairly and write a short description of how it would feel if an AI did the same
• Make a simple two‑column list with “Fair AI Uses” on one side and “Risky AI Uses” on the other, and add at least three examples to each