AI bias occurs when systems produce systematically unfair outcomes for certain groups — often reflecting and amplifying biases already present in training data or system design.

How Bias Enters AI Systems

• Training data bias: Historical data encodes historical discrimination
• Label bias: Human annotators bring their own biases to labeling
• Sampling bias: Underrepresented groups get worse model performance
• Feedback loops: Biased outputs create biased future training data

Real Example

Amazon scrapped an AI recruiting tool in 2018 after discovering it systematically downgraded resumes from women because it was trained on historical hiring data dominated by male candidates.

Reference:

Algorithmic Justice League

https://www.ajl.org/