Guide-linked quiz · Medium Level
Base Rates and False Positives in AI Detection Quiz
Works through prevalence, sensitivity, false-positive rates, predictive value, thresholds, and why detector flags require contextual review.
Question 1 of 8
A condition is rare in a population. Why can a detector with a modest false-positive rate produce many false alarms?
Preview every question8 questions
Answers and explanations appear as you take the quiz.
- A condition is rare in a population. Why can a detector with a modest false-positive rate produce many false alarms?
- There are many more unaffected cases to generate false positives
- Sensitivity becomes zero
- Every negative case is counted as positive
- Prevalence cancels from the calculation
- In the guide's example, 1 percent prevalence among 1,000 items means how many actual positive items?
- 50
- 10
- 90
- 100
- What does sensitivity measure?
- The share of all flags that are correct
- The condition's prevalence
- The share of actual positive cases that the detector flags
- The share of negative cases incorrectly flagged
- Which question does positive predictive value answer?
- Among positive cases, what share are flagged?
- How common is the condition before testing?
- How many thresholds exist?
- Among flagged cases, what share are truly positive?
- Why should a benchmark's evaluation population be compared with the deployment population?
- Different prevalence can change the meaning of a positive flag
- Benchmarks never use labels
- It makes sensitivity unnecessary
- The detector will otherwise refuse to run
- What commonly happens when a detector threshold is lowered?
- It eliminates every false positive
- It may catch more positives while producing more false positives
- It leaves sensitivity unchanged by definition
- It changes prevalence
- For a consequential decision, how should a detector flag be used?
- As conclusive proof of the condition
- As a replacement for an appeal process
- As a prompt for additional evidence and review
- As a reason to ignore context
- Why report an evaluation threshold with error rates?
- Because thresholds determine prevalence
- Because sensitivity is unrelated to thresholds
- Because a single accuracy score contains all details
- Because the threshold defines the operating tradeoff between misses and false alarms
Keep testing yourself
More quizzes picked for your level.