Question: Explain how the practical uses of "bias" and "avoidable bias" in a machine learning strategy text differ from their formal statistical meanings. Why are the practical definitions preferred in that setting?

Sample answer: In the practical machine learning sense, bias refers to the part of a model's error that comes from not fitting the data well enough, and avoidable bias is the portion of that error that could be reduced by improving the model or training process. In formal statistics, bias has a more technical meaning: it is the gap between an estimator's expected value and the true quantity being estimated. The practical definitions are used because they help engineers decide what to improve next, such as model size, features, or optimization, instead of forcing them to think in abstract statistical terms.

Key points:

• Practical bias is an action-oriented notion of model error
• Avoidable bias is the reducible part of that error
• Formal statistical bias concerns estimator expectation
• The practical wording is chosen to guide model improvement decisions

Rubric: Full credit for distinguishing the practical terms from the formal statistical ones and explaining that the practical wording is used because it is more useful for improving models.