AI Risk Assessment, Thresholds, and Treatment AAISM Practice Questions (Page 7)

Questions 31–35

1. 31expert · hard

   An insurance company is assessing the risk of an AI underwriting model that could produce biased outcomes. They have limited historical data on bias incidents but have expert opinions from data scientists and compliance officers. The company needs to prioritize this risk against other AI risks. Which analysis approach is most appropriate given the data availability?

   • AUse qualitative analysis based on expert judgment to rate likelihood and impact, then prioritize using a risk matrix.
   • BAssign a 'medium' rating to all risks to avoid bias in the assessment.
   • CDefer the risk assessment until more data is collected over several years.
   • DUse quantitative analysis by estimating the probability of bias from similar industry incidents.
   Reveal Solution Discussion Select an answer first
2. 32foundation · easy

   Which sequence best represents the systematic process of AI risk assessment?

   • ARisk treatment, risk analysis, risk identification
   • BRisk analysis, risk evaluation, risk identification
   • CRisk identification, risk analysis, risk evaluation
   • DRisk evaluation, risk identification, risk analysis
   Reveal Solution Discussion Select an answer first
4. 33foundation · easy

   What is residual risk in the context of AI risk management?

   • AThe risk that is identified during the initial assessment
   • BThe risk that is transferred to a third party
   • CThe risk that is completely eliminated by treatment
   • DThe risk that remains after risk treatment measures have been implemented
   Reveal Solution Discussion Select an answer first
5. 34application · medium

   A healthcare AI system that predicts patient deterioration has a potential risk of false negatives. The risk team must analyze this risk to decide if treatment is needed. They have historical data on false negative rates and the impact of missed detections. Which analysis method is most appropriate?

   • AA simple binary classification: if the risk is mentioned in the news, treat it; otherwise, ignore it.
   • BQuantitative analysis using the false negative rate and the cost of a missed detection to calculate expected loss.
   • CDefer analysis until the system is in production and real-world performance data is available.
   • DQualitative analysis based on the team's expert opinion that false negatives are 'high risk'.
   Reveal Solution Discussion Select an answer first
6. 35application · medium

   A bank's AI credit-scoring model has a risk of bias against certain demographics. The risk is rated as 'medium likelihood, medium impact.' The bank's risk threshold for AI models is set at 'high' severity. According to the threshold, this risk is below the threshold. What is the appropriate action?

   • AAccept the risk because it falls below the threshold, but document the decision and monitor for changes.
   • BReclassify the risk as 'high' to force treatment.
   • CTransfer the risk to a third-party auditor to avoid responsibility.
   • DTreat the risk immediately because any bias risk is unacceptable.
   Reveal Solution Discussion Select an answer first