Selection Bias

A bias introduced when the training data is not representative of the target population, leading to models that perform unevenly across different groups or scenarios.

When your training data doesn't represent the real world. If you train a medical AI only on data from young adults, it may perform poorly on elderly patients.

Selection bias is a fundamental risk that your AI governance framework must address at the data collection stage. If training data does not represent your actual user population, the AI will not serve all users fairly.