Policy Design CRAGE Practice Questions (Page 7)

Questions 31–35

1. 31application · medium

   A city government is developing an AI policy for a predictive policing system. Community advocates have expressed concerns about bias, while police leadership wants operational flexibility. The policy team must design a process that incorporates diverse perspectives while maintaining the ability to finalize the policy. Which stakeholder engagement method best balances these needs?

   • AHold a series of public town halls and advisory committee meetings, then synthesize the feedback into a draft policy for formal approval.
   • BInvite only internal law enforcement stakeholders to design the policy, since they understand operational needs best.
   • CDefer all policy decisions to an external consulting firm and publish their recommendations unchanged.
   • DConduct a single online survey and use the results as the final policy without further review.
   Reveal Solution Discussion Select an answer first
2. 32expert · hard

   A hospital is implementing an AI system that predicts patient deterioration. The system has shown high accuracy in trials, but there are concerns about alert fatigue among nurses, which could lead to ignored alerts. The policy team must design a risk mitigation strategy that balances patient safety with clinician workload. Which strategy is most effective?

   • ASet a high threshold for alerts to reduce false alarms, and provide additional training for nurses on when to escalate.
   • BDisable alerts for predictions below a certain probability to eliminate false alarms entirely.
   • CImplement a tiered alert system where high-risk predictions trigger immediate alerts and lower-risk predictions are summarized in a daily report.
   • DRequire nurses to acknowledge every alert, regardless of severity, to ensure accountability.
   Reveal Solution Discussion Select an answer first
4. 33expert · hard

   A global tech company is designing an AI governance policy for its facial recognition product. The product is sold to law enforcement agencies in several countries, including EU member states where the EU AI Act may classify it as high-risk. The company also faces pressure from human rights NGOs to halt sales. The policy team must balance business interests, regulatory compliance, and ethical concerns. Which policy approach is most balanced?

   • AContinue sales without additional restrictions, relying on customers to ensure compliance.
   • BLobby regulators to weaken the EU AI Act requirements for facial recognition to reduce compliance burden.
   • CStop all sales of the product immediately to avoid any ethical or legal risk.
   • DAdopt a policy that restricts sales to customers who demonstrate compliance with the EU AI Act and implement a human-rights impact assessment for each deployment.
   Reveal Solution Discussion Select an answer first
5. 34application · medium

   A multinational bank is designing an AI governance policy for a credit-scoring model that will be deployed in EU member states. The policy team wants to ensure the model's decisions can be explained to customers and that the policy aligns with the EU AI Act's risk-based requirements. Which approach best satisfies these needs?

   • AImplement a policy that treats all AI models equally with a single set of requirements, avoiding risk classification to simplify compliance.
   • BDelegate all policy decisions to the legal department to ensure GDPR compliance, focusing only on data protection impact assessments.
   • CRequire the model to be classified as high-risk and mandate that the policy includes documentation of training data, model performance metrics, and a human-review process for adverse decisions.
   • DAdopt a general transparency policy that requires the AI team to publish model accuracy scores on the company intranet without specifying risk classification.
   Reveal Solution Discussion Select an answer first
6. 35application · medium

   A multinational bank is designing an AI governance policy for a credit-scoring system that will be deployed in both the EU and the UK. The policy team wants to ensure the policy is enforceable and aligned with regulatory expectations. Which policy design approach best satisfies the accountability principle while maintaining regulatory alignment?

   • ADelegate all accountability to the external vendor that built the model, since the bank only uses the output.
   • BDefine a single global policy that assigns a named human owner for the AI system's decisions and includes a documented escalation path for disputes.
   • CCreate separate policies for each jurisdiction and require the data science team to interpret them independently during model development.
   • DPublish the policy on the intranet and require all employees to acknowledge it annually, without specifying decision ownership.
   Reveal Solution Discussion Select an answer first