Section 3 – Practice Activities

For this section, your project is to deeply explore fairness and impact by creating a story and review about one AI system that affects real people.

Start by choosing a situation where AI could influence people differently. It might be AI used to pick students for a special program, to recommend news and videos to teenagers, to help doctors sort patients by urgency, or to score job applications. On paper or in a document, write a story of one to two pages from the point of view of at least two different characters who are affected by this system. Show what the AI does in their lives, what they like about it, and what feels worrying or unfair.

When your story is finished, add three short sections on a new page. Title the first section “Possible Unfairness” and list at least three ways the AI might treat people differently—for example, because of biased data, language differences, or unequal access to devices. Title the second section “Questions for Designers” and write at least five questions you would want to ask the people who built this AI, such as what data they trained it on, whether they tested it on different groups, and whether people are allowed to appeal decisions. Title the third section “Changes to Make It Fairer” and suggest at least three changes that could reduce unfairness, like collecting more balanced data, requiring human review for important decisions, or providing clear explanations to users.

If you have access to Hackidemia’s speech-to-text tools, you can turn part of your story into an audio experiment. Visit: https://hackidemia.com/
Read the same paragraph of your story out loud in two different ways: once in your normal voice and once with a different speed or accent. Compare how accurately the system transcribes each version. Notice whether it handles some ways of speaking better than others. Then, add a short paragraph to your “Possible Unfairness” section about how speech recognition AI might accidentally favor some speakers over others, and what could be done about that.

If you have access to Teachable Machine, you can also test fairness in a simple visual model. Use: https://teachablemachine.withgoogle.com/
Create a model that recognizes a basic hand gesture or a simple object. Then ask several people with different skin tones, hand sizes, or backgrounds to test it. Record any differences in accuracy that you see and update your “Changes to Make It Fairer” section with ideas on how to improve the model, such as collecting more diverse examples or changing the background.