AI School › The dangers · Lesson 22 of 130

⚖️ Bias: when AI is unfair

If the data is skewed, AI learns prejudice. And repeats it with confidence.

Remember that AI learns from data? Here is the problem: if the data contains injustice, the AI learns it as if it were normal. If it has seen almost exclusively photos of male scientists, it will draw male scientists. If certain neighbourhoods or people appear in the data in a negative light, the AI will treat them worse. This flaw is called bias, meaning prejudice.

It is a serious problem because AI is also used for big decisions about adults: who gets hired, who gets a loan. That is why we need more balanced data, constant checks and people who keep watch. You can do your part too: when an answer seems unfair to someone, say it. Machines repeat; people can correct.

🎒 Try it in class (or at home)

Experiment: ask an AI to describe "a genius", "a strong person", "a champion". What do they look like? What names? Count together: do the answers represent all the kinds of people you know?

← Lesson 21: Scams and strange messages: spotting them · Lesson 23: Screens and chatbots: time slipping away →

📱 Download the Android app (beta) iOS coming soon

Say what you mean. Get what you need.

Grace Certified, the AI coach that trains and certifies your prompt engineering, by Agora Intelligence.