Conduct Risk Assessment(s) (e.g., Security, Safety, Ethics) PMI-CPMAI Practice Questions (Page 4)

Questions 16–20

1. 16foundation · easy

   In the context of AI risk assessment, what does 'failure mode identification' primarily aim to do?

   • AAssign a probability to each potential failure
   • BDetermine the financial cost of a system failure
   • CEliminate all potential failures from the system
   • DIdentify the ways in which an AI system can fail and the potential safety implications
   Reveal Solution Discussion Select an answer first
2. 17foundation · easy

   An AI hiring tool is found to consistently rank candidates from a certain demographic group lower than others with similar qualifications. Which ethical risk does this primarily represent?

   • ABias and unfairness
   • BLack of accountability
   • CData privacy violation
   • DLack of transparency
   Reveal Solution Discussion Select an answer first
4. 18application · medium

   A company is developing an AI system to recommend personalized investment portfolios for its clients. The system is based on a reinforcement learning algorithm that is designed to maximize returns. During a risk assessment, the team is identifying potential failure modes. Which of the following is the most likely failure mode for this type of AI system?

   • AThe system may be unable to process large volumes of data, leading to slow response times.
   • BThe system may be vulnerable to adversarial attacks that cause it to recommend high-risk investments.
   • CThe system may learn to maximize short-term returns at the expense of long-term client goals, such as retirement savings.
   • DThe system may be biased against certain demographic groups, leading to unfair investment recommendations.
   Reveal Solution Discussion Select an answer first
5. 19application · medium

   A financial services firm deploys a machine learning model that approves or rejects loan applications. The model is trained on historical data and is exposed to the public through an API. During a review, the data science team discovers that an attacker could submit specially crafted inputs that cause the model to consistently approve loans for applicants who should be rejected. The team needs to address this vulnerability before the next release. What should the team do first?

   • AConduct a formal cybersecurity vulnerability assessment to characterize the attack surface, the specific model vulnerabilities, and the potential business impact.
   • BReplace the current model with a simpler, rule-based system that is not susceptible to adversarial inputs.
   • CImmediately shut down the public API to prevent any further exploitation of the model.
   • DImplement adversarial training by augmenting the training dataset with adversarial examples and retraining the model.
   Reveal Solution Discussion Select an answer first
6. 20expert · hard

   A company is developing an AI system for hiring. The system is trained on historical hiring data and is used to screen job applicants. An internal audit reveals that the system is biased against women. The company is committed to developing a mitigation strategy. Which of the following is the most effective mitigation strategy?

   • AUse a fairness-aware algorithm that explicitly optimizes for equal outcomes across all demographic groups.
   • BUse a different AI algorithm that is known to be less susceptible to bias.
   • CAdd more data about women to the training set to balance the representation.
   • DRemove the gender attribute from the training data and retrain the model.
   Reveal Solution Discussion Select an answer first