AI for Forensic Accountants Quiz

Answers and explanations appear as you take the quiz.

1. A first-two-digit Benford test on invoices shows a spike at 49, and the company's approval limit is $5,000. How should a forensic accountant treat this result?
   • As proof that the approver committed fraud
   • As a lead suggesting purchases may have been split to stay under the limit, which needs follow-up
   • As a sign the dataset is too small for Benford analysis
   • As evidence that Benford's law never applies to invoices
2. Which of these datasets is least suitable for Benford's law analysis?
   • Vendor invoice amounts ranging from $10 to $500,000
   • Employee expense claims across many sizes
   • Sequentially assigned invoice numbers
   • Sales totals by customer over a year
3. Under Benford's law, roughly how often should the leading digit be 1 in a conforming dataset?
   • About 30 percent of the time
   • About 11 percent of the time
   • About 46 percent of the time
   • About 5 percent of the time
4. Why do practitioners often prefer mean absolute deviation over chi-square when testing Benford conformity on very large datasets?
   • Mean absolute deviation needs no computer to calculate
   • Chi-square only works on second-digit tests
   • Mean absolute deviation automatically identifies the fraudster
   • Chi-square treats trivial deviations as significant when the dataset is large
5. Link analysis reveals that a supplier is registered at a company director's home address. What does this most directly suggest?
   • The supplier's invoices follow Benford's law
   • The director has a conflict of interest that was properly disclosed
   • A possible undisclosed related party relationship
   • A problem with the tracing method chosen
6. In fund tracing through commingled accounts, what decides whether a method such as the lowest intermediate balance rule is used?
   • The default setting of the tracing software
   • A legal choice about which tracing convention applies
   • The number of transactions in the account
   • Whether the data conforms to Benford's law
7. What do Federal Rule of Evidence 702 and the Daubert standard require of expert testimony in US federal courts?
   • That every finding be produced by machine learning
   • That the expert use only Benford's law
   • That the analysis cover the entire population of transactions
   • That the testimony rest on reliable methods that were properly applied
8. When resolving entities for link analysis, which shared attribute gives the strongest evidence that two records belong to the same party?
   • A shared bank account number
   • A shared surname
   • A shared city
   • A shared industry code