A risk committee is evaluating whether to deploy an AI system for automated debt collection communications. The system was trained primarily on historical communications from accounts that were successfully resolved. What dataset bias risk does this create?
Select an answer to reveal the explanation.
Short Explanation and Infographic
Here's the deal — a is correct because training exclusively on successfully resolved accounts creates survivorship bias — the model learns what worked for cases where collection succeeded but has no exposure to accounts where standard approaches failed, potentially rendering it less effective for harder collections. B is a volume concern, not the specific bias type.
Full explanation below image
Full Explanation
A is correct because training exclusively on successfully resolved accounts creates survivorship bias — the model learns what worked for cases where collection succeeded but has no exposure to accounts where standard approaches failed, potentially rendering it less effective for harder collections. B is a volume concern, not the specific bias type. C is a human cognitive bias in development, not a dataset bias. D is a temporal generalization concern, not the primary bias identified.