Chapter 6: Detecting ML Bias and Explaining Models with SageMaker Clarify

Machine learning (ML) models are increasingly being used to help make business decisions across industries, such as in financial services, healthcare, education, and human resources (HR), thanks to the automation ML provides, with improved accuracy over humans. However, ML models are never perfect. They can make poor decisions—even unfair ones if not trained and evaluated carefully. An ML model can be biased in a way that hurts disadvantaged groups. Having an ability to understand bias in data and ML models during the ML life cycle is critical for creating a socially fair ML model. SageMaker Clarify computes ML biases in datasets and in ML models to help you gain an understanding of the limitation of ML models so that you can take appropriate action to mitigate these biases.

ML models have long been considered as black box operations because it is rather difficult to see how a prediction is made. SageMaker...