AI in Credit Underwriting Quiz

Answers and explanations appear as you take the quiz.

1. What does an AI credit-underwriting model primarily try to predict?
   • The probability that the borrower will default
   • The borrower's favorite bank
   • The stock market direction
   • The borrower's age exactly
2. What is 'proxy discrimination' in credit models?
   • Using a backup server
   • A feature standing in for a protected attribute like race
   • Hiring a proxy to apply for loans
   • Encrypting customer data
3. What does an adverse-action notice communicate to a credit applicant?
   • A guarantee of future loan approval
   • A credit score that proves discrimination
   • That the creditor took an adverse action on the credit application
   • A list of every variable in every model
4. Who can AI underwriting potentially help that traditional scorecards often reject?
   • Only the wealthiest borrowers
   • Banks themselves
   • Government regulators
   • 'Thin-file' applicants with little traditional credit history
5. Why are tools like SHAP used in credit underwriting?
   • To attribute a decision to specific features for explanations
   • To speed up the internet
   • To set the prime interest rate
   • To detect computer viruses
6. What is the central tension in using more data for underwriting?
   • More data always lowers accuracy
   • More data can improve accuracy but may hide discrimination
   • Data is illegal to use
   • Data makes loans free
7. When a creditor supplies a statement of reasons under Regulation B, what must those reasons describe?
   • A generic statement that an algorithm denied credit
   • A complete source-code listing
   • Specific principal reasons that accurately describe factors actually used in the decision
   • A promise that the applicant will later be approved
8. Why can an alternative-data source raise fairness questions in underwriting?
   • It is always illegal to use any nontraditional data
   • It guarantees a higher credit score
   • It eliminates the need for validation
   • It may act as a proxy for protected characteristics or reflect uneven coverage