Level 1 · User
AI has biases too
It learned from human text, our biases included.
If, in the text it learned from, the doctor is almost always a man and the nurse a woman, AI tends to repeat that. It isn't malicious: it reflects what it saw.
It shows in details: who it pictures in an image, what tone it uses with certain names, or which profile it recommends for a role.
Try it today
Ask an image AI for "a team of engineers". Look at who appears, then ask again specifying diversity.
Test yourself
What's the best practice?
- Use AI to decide who to hire
- Ask for several options and let a person decide
- Never use AI
See the answer
Ask for several options and let a person decide. AI helps widen the options. Decisions about people are made by people.
If you lead a team
If AI touches decisions about people (hiring, credit, reviews), you need human review and a bias test before launch.