Free GARP RAI Exam Questions

Question 1

An insurance company uses two different AI models for policy approval, one for Group A and another for Group B. The confusion matrices for each group are as follows:Group A Confusion Matrix:True Positives: 50False Positives: 10False Negatives: 40True Negatives: 90Group B Confusion Matrix:True Positives: 30False Positives: 20False Negatives: 70True Negatives: 80What fairness technique should the company use to balance the performance between the two groups?

A : Equalize the number of true positives for both groups.

B : Adjust thresholds to achieve similar false positive rates for both groups.

C : Apply a higher penalty for false negatives in Group A.

D : Implement demographic parity by ensuring equal acceptance rates for both groups.

Answer: D

Question 2

In an effort to understand why certain loan applications were denied, a bank deploys LIME (local interpretable model-agnostic explanations). What is the primary advantage of using LIME?

A : LIME trains a new model for each decision made by the AI

B : LIME provides an approximate explanation of a complex model’s behavior in a local context

C : LIME permanently alters the original model to make it interpretable

D : LIME identifies the most accurate model among several candidates

Answer: B

Question 3

You are given the following feature vector and statistical parameters:Feature Vector: [1, 3.6, 9, 10]Mean: 5Variance: 2.89What is the standardized feature vector?

A : [−2.35,−0.82, 2.35, 2.94]

B : [−2.34,−0.81, 1.81, 1.97]

C : [−1.81,−0.81, 2.34, 2.76]

D : [−2.34,−0.81, 2.34, 3.76]

Answer: A

Question 4

A financial institution is creating a model inventory. Which of the following should be included to ensure comprehensive governance?

A : Only the model’s purpose and validation report.

B : Model owner, but not developer information, for clarity.

C : A description of model assumptions but not its risk classification.

D : Model owner, purpose, methodology, assumptions, limitations, and risk tier.

Answer: D

Question 5

An insurance company uses two different AI models for policy approval, one for Group A and another for Group B. The confusion matrices for each group are as follows:Group A Confusion Matrix:True Positives: 50False Positives: 10False Negatives: 40True Negatives: 90Group B Confusion Matrix:True Positives: 30False Positives: 20False Negatives: 70True Negatives: 80What fairness technique should the company use to balance the performance between the two groups?

A : Equalize the number of true positives for both groups.

B : Adjust thresholds to achieve similar false positive rates for both groups.

C : Apply a higher penalty for false negatives in Group A.

D : Implement demographic parity by ensuring equal acceptance rates for both groups.

Answer: D