A risk described as model users not understanding underlying assumptions is best captured by which term?

A risk described as model users not understanding underlying assumptions is best captured by which term?

• A. Overfitting
• B. Lack of Comprehension
• C. Data Drift
• D. Model Miscalibration

Answer: (B)

The main idea here is the risk that users of a model don’t grasp the assumptions that underlie how the model works. When people don’t understand these assumptions, they may misinterpret outputs, misapply the model, or overtrust or underutilize its results. That gap in understanding is precisely what “Lack of Comprehension” describes. Overfitting is about the model learning noise in the training data and not generalizing well to new data, a property of the model itself rather than how well users understand its underpinnings. Data drift refers to changes in the input data distribution over time, which affects performance but again concerns data dynamics, not user comprehension of assumptions. Model miscalibration involves the predicted probabilities not matching actual outcomes, a technical calibration issue, not the user’s grasp of how the model relies on its assumptions. So while all are important, they don’t capture the human-facing risk of not understanding the underlying assumptions as directly as lack of comprehension does.

The main idea here is the risk that users of a model don’t grasp the assumptions that underlie how the model works. When people don’t understand these assumptions, they may misinterpret outputs, misapply the model, or overtrust or underutilize its results. That gap in understanding is precisely what “Lack of Comprehension” describes.

Overfitting is about the model learning noise in the training data and not generalizing well to new data, a property of the model itself rather than how well users understand its underpinnings. Data drift refers to changes in the input data distribution over time, which affects performance but again concerns data dynamics, not user comprehension of assumptions. Model miscalibration involves the predicted probabilities not matching actual outcomes, a technical calibration issue, not the user’s grasp of how the model relies on its assumptions. So while all are important, they don’t capture the human-facing risk of not understanding the underlying assumptions as directly as lack of comprehension does.