The Fairness Test: An Unplugged Activity That Reveals Hidden Bias

You’ll need old magazines, scissors, or printed photos. About twenty pictures of people — different ages, different skin tones, different clothes. You’ll also need a simple question: “Who looks like a teacher?”

What happened when I tried it

My daughter was eight. She sorted the pictures quickly. White women in glasses went into the “teacher” pile. A man in a turban went into the “not teacher” pile. A woman with a visible disability went into the “not teacher” pile. Every child — every single one — went into “not teacher.”

When I asked why, she said, “Kids can’t be teachers.”

She wasn’t being cruel. She was reflecting back the world she’d seen.

The three questions that change everything

1. Who did you leave out?
2. Why do you think you left them out?
3. If a robot learned from your choices, who would it think a teacher looks like?

After those three questions, my daughter rebuilt her piles. She added the woman with the disability. She added the man in the turban. She added the children. “They could be teachers one day,” she said. “Or they could teach me something.”

What this teaches about AI

When you build an AI, you train it on data — images, text, numbers. If that data is missing whole groups of people, the AI will be biased. Not because anyone programmed it to be unfair, but because it never saw enough examples. The Fairness Test makes that abstract concept tangible, emotional, and unforgettable.

From table‑top to real life

In HiKIDAI’s AI Fairness Detective, children do this same work — they examine training data, spot missing groups, and re‑train the AI. They learn that fairness is something you build, not something you hope for.