Quiz: Where Bias Enters: Data, Labels, Features and Deployment

Bias in AI systems often originates from the data used to train them, which may reflect historical patterns of discrimination.

Labels in AI systems can reflect human biases or cultural assumptions that become embedded in machine learning models.

Feature selection processes in AI can unintentionally amplify demographic disparities through proxy variables.

The performance of AI systems can vary significantly across demographic groups when they are deployed in real-world environments.

Monitoring AI systems during deployment helps identify when bias emerges or increases through ongoing evaluation.