Question: A team wants to estimate apartment rent in Seattle using unit size and bedroom count. They also have rent records from Cleveland and want to combine the two sources. Explain when the outside dataset is inconsistent with the Seattle task and why combining them can weaken the learner.

Sample answer: An outside dataset is inconsistent when the same feature pattern is associated with different outcomes in different sources. For example, a 900-square-foot, two-bedroom apartment may rent for much more in Seattle than in Cleveland. If a model is trained on both datasets together, it sees one input paired with very different rent values. That conflict makes it harder to learn the Seattle-specific relationship between features and rent, so performance on the target task can fall.

Key points:

• Inconsistency means one set of features does not correspond to one stable outcome across sources.
• The same apartment size can imply very different rents in Seattle and Cleveland.
• Combining the datasets can give the learner mixed signals.
• Conflicting auxiliary records can reduce accuracy on the target city.

Rubric: A strong response will define inconsistency as a mismatch in the feature-to-outcome relationship, apply it to a Seattle-versus-Cleveland housing example, and explain that the model receives conflicting training signals that hurt target-task learning.