Module 12 Knowledge Check
5 questions. Pass mark 4 out of 5. Answer every question before checking the answer key below, then retry after reading the feedback.
1. The scope should define the model by…
- A. its architecture
- B. the decision it supports
- C. its accuracy target
- D. its training data size
2. A class with five examples should be…
- A. measured and reported
- B. reported as unmeasured
- C. excluded from the dataset
- D. oversampled to match the others
3. The model should be reported against…
- A. other architectures
- B. the majority baseline and a trivial baseline
- C. the previous version only
- D. published benchmarks
4. The deployment decision should be made…
- A. on the overall metric
- B. per slice
- C. on the validation set
- D. by the highest-scoring threshold
5. In an honestly measured model card…
- A. permissions outnumber prohibitions
- B. prohibitions usually outnumber permissions
- C. neither is listed
- D. only intended use is listed
---
Answer Key and Explanations
Check these only after attempting every question.
1. B - the decision it supports
That determines the metric, the threshold and the refusal conditions.
2. B - reported as unmeasured
Five examples support no performance claim, and including it in an overall figure hides that.
3. B - the majority baseline and a trivial baseline
Accuracy can move 0.7 points while the metric that matters moves 62 - the baselines are what make that visible.
4. B - per slice
An overall recall of 62% can average over slices differing by 25 points, and the weak ones are usually smallest.
5. B - prohibitions usually outnumber permissions
Unvalidated slices, unvalidated devices and prohibited purposes accumulate faster than approved uses.
Practical Check
Apply this module to your own work: complete the module activity for *Capstone: Responsible Vision Workflow*, then write one sentence naming what your result shows and one naming what it does not.
Strong Answer Pattern
A strong answer names the task, the evidence used, the check performed, and the remaining limitation. It avoids "proved", "guaranteed", or "always" unless the evidence genuinely supports it.