◆ Definition
Bias (in AI)
Systematic unfairness in model behavior, usually inherited from patterns in training data. Shows up as skewed outputs for certain groups, dialects, or contexts — a data and evaluation problem more than a code problem.
Explained for kids
When AI is unfair because of what it learned. If a model only ever saw pictures of golden retrievers, it might decide a chihuahua isn't a dog. Nobody meant to be unfair — the examples were lopsided.
Learn more in these free lessons
- AI Ethics & Bias · AI Essentials
- Automation Bias and Deployment Design · AI for Healthcare
- Bias: Where It Actually Shows Up · AI for School Students
- Dataset Bias and Documented Harms · AI for Healthcare
- Leakage and Look-Ahead Bias · AI for Finance
- LLM-as-Judge, Done Properly · Does Your AI Actually Work?