Reference · Glossary

Bias

Systematic **unfair skew** in AI outputs — often reflecting biased training data or design choices that hurt groups or scenarios.

When to use

Discussing fairness, hiring tools, healthcare triage, content moderation, and dataset audits.

When not to

Every model mistake is not bias — some are random errors or missing context. Investigate patterns across groups.

Example

A resume screener consistently ranks identical CVs lower when names suggest certain demographics — a bias red flag.