F27 · Topic 5.3
Recommendation bias investigation
1. Predict first
Group A has 20 cases and 8 errors; group B has 80 cases and 4 errors. Predict both accuracies and the overall accuracy.
Change it · step through · explain
Change an input, predict what will happen, then advance one step. Read why the state changes, reset and experiment. This fixed teaching model does not save progress or award XP.
Ready · Make a prediction
- Input 120,80Waiting
- Input 28,4Waiting
Ready to experiment
Why: Predict the first change, then advance one step.
Try this: Change group sizes while keeping very different error rates. Does a high overall accuracy hide a gap?
2. Build your procedure
Compare the baseline with a proposed error count 4 for A and 8 for B, holding group sizes fixed. Explain what improves, what worsens and what cannot be concluded from rates alone.
Use AP CSP pseudocode, a language you know, or precise plain language. Your code is not executed here. Use the classroom assignment to save your response.
3. Check boundaries
| Input | Expected | Why this case |
|---|---|---|
| Sizes 20,80; errors 8,4 | Accuracies 60%,95%; overall 88% | Use each group’s denominator. |
| Sizes 20,80; errors 4,8 | 80%,90%; overall 88% | Same aggregate, different distribution. |
| An error count above its group size | Reject inconsistent input | An error rate cannot exceed the stated number of cases. |
4. Explain and reflect
Name two development choices to investigate and one privacy constraint when evaluating unequal results.
Model limits: Aggregate synthetic counts only. The tool neither infers real protected traits nor declares a classifier legally or ethically fair.