引導式測驗 · 中等 等級
預測性警務演算法測驗
測試預測性警務概念:基於地點的預測與基於個人的預測、有偏見的記錄、回饋循環、影響評估、審計和法律
問題 1 的 8
A place-based system marks blocks with higher predicted incident counts. What does that output represent?
預覽每個問題8 問題
當您參加測驗時,就會出現答案和解釋。
- A place-based system marks blocks with higher predicted incident counts. What does that output represent?
- Proof that every person on the block is suspicious
- A judicial finding that a crime occurred
- A forecast about recorded incidents under the model’s data and assumptions
- A guarantee that more patrols will prevent crime
- How can increased patrols create a feedback loop in a predictive system?
- More deployment can generate more recorded activity, which the model may treat as evidence for future deployment
- Patrols automatically make old records disappear
- More deployment removes bias from the training set
- An algorithm stops updating whenever officers make contact
- A person-based tool flags an individual as high risk. What does the score authorize by itself?
- An arrest because the score is a probability
- A search whenever the person enters a forecasted area
- A presumption that the person will offend
- No stop, search, or arrest; ordinary legal standards still apply
- Before deploying a predictive-policing system, which DOJ-recommended step is relevant?
- Let the vendor select the only evaluation data
- Complete an impact assessment and test the system before use
- Wait until a serious error occurs before auditing
- Use overall accuracy as the entire review
- A database includes people whose records no longer meet the tool’s eligibility criteria. Which control addresses this?
- Increase the model’s number of features
- Hide record criteria from auditors
- Routinely audit the database and correct or remove ineligible records
- Treat every historical entry as valid
- A model predicts reported calls well but has no measure of unreported events. What limitation follows?
- It has proved equal performance across neighborhoods
- It establishes the true crime rate
- It can be used without local validation
- Its target is recorded calls, which may not represent all underlying incidents
- Why should an independent auditor be able to examine more than a vendor’s headline accuracy score?
- To assess reliability, bias, inputs, and performance under the intended deployment conditions
- To guarantee that every future prediction is correct
- To let the auditor set arrest standards
- To replace public policy review with a technical test
- Which outcome is most informative when judging a new deployment?
- Count how many alerts the system produces
- Compare public-safety results and community burden with a credible alternative
- Measure whether officers open the map
- Report only whether predictions match past training data