Which concept describes the issue where using flawed proxies (like healthcare spend as a proxy for sickness) can bias results toward wealthier groups?

Which concept describes the issue where using flawed proxies (like healthcare spend as a proxy for sickness) can bias results toward wealthier groups?

• A. Problem Specification
• B. Representation Bias
• C. Imbalanced Dataset
• D. Explainability

Answer: (A)

The central idea is how you define the problem and what you use as the target signal. If sickness is inferred from a flawed proxy like healthcare spend, the proxy carries not just information about illness but also information about wealth and access to care. Wealthier individuals often have higher healthcare spending regardless of sickness, so the model learns patterns tied to wealth rather than true illness. That makes the results biased toward wealthier groups because the target signal itself is mis-specified by the proxy. This is different from representation bias, which comes from the training data not reflecting the real population, and from an imbalanced dataset, which concerns the relative frequencies of outcomes rather than how the outcome is measured. Explainability touches on understanding model decisions, not how the problem was framed or what variable was chosen as the measure of sickness.

The central idea is how you define the problem and what you use as the target signal. If sickness is inferred from a flawed proxy like healthcare spend, the proxy carries not just information about illness but also information about wealth and access to care. Wealthier individuals often have higher healthcare spending regardless of sickness, so the model learns patterns tied to wealth rather than true illness. That makes the results biased toward wealthier groups because the target signal itself is mis-specified by the proxy.

This is different from representation bias, which comes from the training data not reflecting the real population, and from an imbalanced dataset, which concerns the relative frequencies of outcomes rather than how the outcome is measured. Explainability touches on understanding model decisions, not how the problem was framed or what variable was chosen as the measure of sickness.