Question 1

A justice system uses an AI model to predict failure-to-appear (FTA) rates in court. To ensure fairness without explicitly using protected group data, what could be a potential unintended outcome?

Options :

A : The model underestimates FTA for all demographics

B : The model’s accuracy decreases for high-risk groups

C : The model’s reliance on correlated variables increases potential bias in predictions

D : The model achieves demographic parity but lacks predictive accuracy

Answer: C

Question 2

An investment firm applies the Monte Carlo method for a finite series of market simulations to calculate expected portfolio returns. The method converges slowly, especially during long simulations. Which of the following describes a primary disadvantage of the Monte Carlo method for this application?

Options :

A : It requires continuous data

B : It can only be used for long-term simulations

C : It can only be used for finite-horizon problems

D : It cannot account for discounting future rewards

Answer: C

Question 3

A financial institution uses an opaque consumer scoring algorithm. What is a key reason for keeping the algorithm confidential?

Options :

A : Ensuring users can easily understand their scores

B : Protecting proprietary information and competitive advantage

C : Allowing competitors to improve similar systems

D : Sharing all code with the public for transparency

Answer: B

Question 4

Which of the following scenarios represents a binary classification problem?

Options :

A : Predicting the probability of different stock prices

B : Forecasting next month’s sales in dollars

C : Determining if a transaction is fraudulent or legitimate

D : Grouping customers into three income tiers

Answer: C

Question 5

During an NLS optimization, the analyst uses a gradient descent algorithm to update model parameters. If the improvement in the objective function falls below a certain threshold, the optimization process stops. What is the purpose of this threshold in NLS optimization?

Options :

A : To avoid overfitting by stopping after a set number of steps

B : To ensure convergence to the exact global minimum

C : To prevent unnecessary computations when further improvement is minimal

D : To improve model interpretability by limiting parameter updates

Answer: C