AI in Insurance Underwriting Quiz

Answers and explanations appear as you take the quiz.

1. In a simplified hypothetical risk model, how is expected loss for one exposure calculated?
   • Claim probability multiplied by claim severity
   • Claim probability plus the policyholder’s ZIP code
   • Severity divided by the number of features
   • Premium multiplied by the number of claims
2. Why might a regulated insurer document model inputs and validation?
   • To guarantee that a model never produces biased outcomes
   • To support governance, accountability and review of AI systems
   • To avoid retaining any information about the model
   • To replace actuarial review in every jurisdiction
3. Under the NAIC model bulletin, who remains responsible for AI-supported insurance decisions complying with applicable law?
   • The software vendor alone
   • The applicant whose data were analyzed
   • The insurer using the AI system
   • The model itself as an independent legal person
4. What fairness risk should be tested when a model uses variables correlated with protected traits?
   • The model becomes unable to estimate any claims
   • Every applicant receives the same premium
   • The model automatically removes all protected-class effects
   • A proxy variable may contribute to disparate outcomes
5. In an underwriting model, what is a feature?
   • An input variable used by the model to estimate a risk-related outcome
   • A legal approval for a premium
   • A final insurance claim payment
   • A guarantee that a prediction is fair
6. What does SHAP provide for a model prediction?
   • A legal approval for the resulting premium
   • Feature-attribution values that explain contributions under the SHAP framework
   • Proof that the model is unbiased
   • A guarantee that a feature caused the outcome
7. Which driving signals does Progressive say its Snapshot program may measure?
   • Medical history, prescriptions, and blood pressure
   • Social-media sentiment and shopping history
   • Time of day, hard braking or acceleration, and amount driven
   • Vehicle paint color, route scenery, and music choice
8. How can aerial imagery support property underwriting in Zurich's Nearmap example?
   • It directly measures a home's indoor air quality
   • It replaces all policyholder disclosures and inspections
   • It predicts the exact date a roof will fail
   • It can provide imagery and roof-condition or maintenance indicators for risk assessment