Case context: A team developing an image recognition app finds through error analysis that 40% of errors are due to dark lighting, and 10% are due to blurry images. However, improving performance on dark lighting requires a complete overhaul of the model architecture, while fixing blurry image errors only requires adding a few data augmentation transformations.

Question: Based on the principles of error analysis, how should the team decide which task to prioritize, considering that the dark lighting category accounts for more errors? What factors beyond the raw error percentages must they evaluate?

Sample answer: The team should not automatically prioritize the dark lighting errors simply because they account for 40% of the mistakes. Error analysis does not provide a rigid mathematical formula for prioritization. Instead, the team must evaluate the expected progress and the amount of work required. Since fixing the blurry images (10% of errors) requires significantly less work (data augmentation) compared to the dark lighting (model overhaul), the team might decide that tackling the blurry images is the better priority for immediate expected progress relative to the effort invested.

Key points:

• Raw error percentages are not a rigid formula for setting priorities.
• The expected progress on the dark lighting vs. blurry images must be assessed.
• The amount of work needed (model overhaul vs. data augmentation) is a deciding factor.

Rubric: The answer should evaluate the scenario by applying the principle that error analysis doesn't yield a rigid priority formula. It must explicitly identify that the team needs to weigh the expected progress and the amount of work needed, concluding that the easier task (blurry images) might take priority despite having a lower error rate.