Lesson 2 of 4 / 7 minute read

Where bias comes from

Bias in AI means a system treats some people, places or ideas unfairly or inaccurately in a consistent way. It usually starts in the data.

If most of the training text comes from a few countries and languages, the model knows those contexts best. Ask about a South African town, a local language or a regional custom and the answers may be thinner or wrong.

Bias also enters through choices people make: what to collect, how to label it and what counts as a good answer. Noticing whose perspective is missing is part of using AI well.

Check your understanding

A model gives weaker answers about your local language. The most likely reason is:

Choose the answer you think is right.