AI Glossary Term
What does “Bias” mean?
A consistent pattern of error or unfairness in data or model behavior.
How Bias works
Example
A face-recognition system trained mostly on lighter-skinned faces makes more errors on darker-skinned faces. Research on commercial systems, such as the Gender Shades study, found uneven error rates across skin tone and gender.
Why it matters
Bias in data or models can lead to consistently worse outcomes for some groups. Noticing it early is cheaper and fairer than fixing harm after deployment.
Common questions
Is all bias bad?
In statistics, a model needs some assumptions, called inductive bias, to learn at all. The harmful kind is bias that systematically disadvantages people or skews results.
How is bias different from random error?
Random errors scatter in different directions and tend to cancel out. Bias is a consistent tilt in one direction.
Can more data fix bias?
Only if the new data is more representative. Simply adding more of the same skewed data can reinforce the problem.