Bài kiểm tra có liên kết hướng dẫn · Trung bình Cấp độ
Bài kiểm tra vụ bê bối về thuật toán thất nghiệp MiDAS của Michigan
Kiểm tra trường hợp MiDAS của Michigan: cờ khác biệt tự động, phát hiện gian lận, phạm vi kiểm toán, thông báo, kháng nghị, độ chính xác của dữ liệu và tác hại của người khiếu nại.
Câu hỏi 1 của 8
A wage cross-match flags a difference between reported pay and benefits. What does that flag establish?
Xem trước mọi câu hỏi8 câu hỏi
Câu trả lời và giải thích xuất hiện khi bạn làm bài kiểm tra.
- A wage cross-match flags a difference between reported pay and benefits. What does that flag establish?
- That the claimant committed fraud
- That every benefit week was overpaid
- That the employer record is correct
- A discrepancy that needs review, not intentional fraud by itself
- What did Michigan’s 2016 MiDAS audit explicitly exclude from its scope?
- A review of system documentation
- A review of MiDAS’s identification of fraudulent claims
- A review of user-access controls
- A review of the audit period
- How should the 2022 Michigan settlement announcement be described?
- A settlement resolving allegations about automated fraud findings and due process
- A court trial proving every MiDAS decision was fraudulent
- A software accuracy certification
- A finding that all benefits were properly paid
- Which notice best helps a claimant contest a wage discrepancy?
- State only that a computer detected fraud
- Give a total debt without a reason
- Identify the employer, benefit week, wage record, calculation, and how to respond
- Direct the claimant to the software vendor
- Why is a human review step important before an automated fraud determination?
- A human reviewer makes all data errors impossible
- Review removes the need for an appeal
- A person’s approval proves the model is unbiased
- A person can examine source records and distinguish an error from intentional misrepresentation
- Which test set is useful for evaluating wage cross-matches?
- Only straightforward single-employer records
- The same cases used to tune the rule, without holdout data
- Cases with corrected wages, delayed bonuses, multiple employers, and partial weeks
- Only claims with confirmed fraud
- Which measure can reveal whether notices are causing unsupported determinations?
- Number of notices sent
- Appeal outcomes and the rate at which findings are reversed
- Number of records processed per hour
- Vendor response time
- An audit tests access controls but not fraud decisions. What conclusion is warranted?
- The tested controls may be assessed, but fraud-determination accuracy remains untested
- The system’s fraud findings are accurate
- The audit establishes claimant intent
- No other review is necessary