Identify Guiding Principles for Responsible AI AI-900 Practice Questions (Page 6)

Questions 26–30

1. 26application · medium

   A bank uses an AI model to approve personal loans. During testing, the model approves 80% of applicants from one postal code but only 60% from another, even though the applicants have similar credit profiles. The bank must address this before deployment. What should the bank do first?

   • AAdd a disclaimer to the loan application stating that approval decisions may vary by location.
   • BDeploy the model and monitor its performance over time to see if the disparity corrects itself.
   • CRetrain the model with more data from the underrepresented postal code and re-evaluate the approval rates across all groups.
   • DRemove the postal code feature from the model to guarantee equal approval rates.
   Reveal Solution Discussion Select an answer first
2. 27application · medium

   A hospital deploys an AI system that predicts patient readmission risk. The system was trained on data from a single region and performs well there, but the hospital plans to use it nationwide. What should the hospital do to ensure the system remains reliable and safe across different regions?

   • ADeploy the model nationwide and rely on continuous monitoring to catch any issues after release.
   • BAdd more features to the model to make it more accurate in all regions.
   • CValidate the model on diverse regional data and retrain or adjust it if performance drops in certain areas.
   • DUse the model only in the original region to avoid any risk of unreliable predictions elsewhere.
   Reveal Solution Discussion Select an answer first
4. 28application · medium · select all that apply

   A company is developing an AI system to screen job applicants. The company wants to ensure the system is fair and inclusive. Which of the following actions should the company take? Select all that apply.

   • AInvolve a diverse group of people in the design and testing of the system.
   • BDocument the system's decision-making process for transparency.
   • CUse only the applicant's resume text without any demographic information.
   • DTest the model's performance across different demographic groups.
   Check answers Discussion Select an answer first
5. 29application · medium · select all that apply

   A company deploys an AI system that controls a robotic arm in a warehouse. The system occasionally makes sudden movements that could injure workers. The company wants to improve safety. Which of the following actions should the company take? Select all that apply.

   • AImplement an emergency stop button that immediately halts the robotic arm.
   • BUse sensors to detect when a worker is too close and slow down or stop the arm.
   • CConduct regular safety testing and maintenance of the AI system.
   • DIncrease the speed of the robotic arm to improve efficiency.
   Check answers Discussion Select an answer first
6. 30expert · hard

   A credit card company uses an AI model to detect fraudulent transactions. The model flags a transaction as fraudulent, but the customer disputes the decision. The company wants to provide a transparent explanation to the customer while also protecting the security of the model. What is the best approach?

   • AShare the full model's decision logic and feature weights with the customer.
   • BRefuse to provide any explanation to protect the model's security.
   • CProvide the customer with a general explanation of the factors that led to the flag, such as unusual location or amount, without revealing specific model details.
   • DTell the customer that the decision is final and they can file a formal complaint.
   Reveal Solution Discussion Select an answer first