An AI system used for medical image analysis begins producing incorrect diagnoses six months after deployment despite unchanged inputs. Root cause analysis reveals the model's training data was collected from a different patient population. This scenario best illustrates which risk?
Select an answer to reveal the explanation.
Short Explanation
You're seeing covariate shift because the model learned from one patient population and is now facing a different input distribution. The diagnostic logic may still be sound, but the images and patient characteristics no longer match what it was trained on.
Full Explanation
The correct risk is covariate shift between training and deployment populations. It happens when production inputs have a different statistical distribution than training inputs, so performance degrades even if the label logic hasn't changed. Here, the model learned from one patient population and now sees another, causing later incorrect diagnoses. Label noise is wrong because it means annotation errors, not population differences. An adversarial backdoor is wrong because it's a deliberate malicious implant in weights. Overfitting to validation data is wrong because it would show poor initial performance, not degradation after months.