A bank uses an AI model to evaluate customer loan applications.

This scenario demonstrates bias associated with unrepresentative training data.

If an AI model is trained exclusively on customers with long credit histories, it may fail to accurately evaluate applicants with limited credit information.

Such applicants fall outside the training population, potentially causing the model to classify them as high risk even when other financial indicators suggest acceptable creditworthiness.

This is an example of selection bias caused by systematically excluding a relevant customer segment from the training dataset.

AWS responsible AI guidance emphasizes using representative datasets that reflect real-world populations and anticipated operating conditions.

The bank should include customers with diverse credit histories in model development and validation datasets.

Fairness assessments should also compare predictions across relevant customer groups and investigate potential disparities.

Amazon SageMaker Clarify can support bias analysis and model explainability.