AI Ethics & Bias
📋 What it is
AI can be biased — unfair to some groups — when its training data reflects human biases or is unbalanced.
🗣️ Coach says
AI learns from data made by people, so it can copy human unfairness. If a hiring AI trains mostly on men’s résumés, it may unfairly rank women lower. The AI isn’t “mean” — it mirrors biased data. Spotting this is a key skill.
🧠 Memory hook
AI mirrors its data — biased or unbalanced data makes a biased, unfair AI.
😂 Giggle
Why did the neural network break up with the spreadsheet?
It needed deeper connections!
😲 Whoa!
A real hiring AI was scrapped after it learned to downgrade résumés that mentioned “women’s” clubs — it had copied a bias hidden in its training data.
✅ Quick check: Why might an AI trained mostly on one group be unfair to others?
Say your answer out loud first — then reveal.
It learned patterns from that group’s data, so it works worse (or unfairly) for people unlike its training data.
Unbalanced training data produces a model biased toward the majority in that data.
🧪 Try it! (2 minutes)
Think of a fairness test: would an AI trained only on YOUR handwriting read a friend’s? (Probably poorly.)