Case context: You are building a message filter for an online marketplace that flags seller messages as risky. Missing a harmful message is bad, but incorrectly flagging a legitimate message also creates problems. Several candidate models are available, and some score well on one measure but poorly on the other.

Question: What evaluation approach should the team use to make model comparison easier, and what is a simple way to turn the two measures into one number?

Sample answer: Use a single combined evaluation score rather than comparing the two measures separately. A simple way to build that score is to compute the arithmetic mean of precision and recall for each model.

Key points:

• Recommend merging precision and recall into one evaluation value.
• Explain that two separate scores can make model selection confusing.
• Use the average of precision and recall as a straightforward computation method.

Rubric: The response must state that precision and recall should be combined into one number to compare models more easily, and it must identify averaging them as a valid way to compute that number.