Task Statement 4.1: Explain the Development of AI Systems That Are Responsible. AIF-C01 Practice Questions (Page 5)

Questions 21–24

1. 21foundation · easy

   A machine learning team is reviewing the key features of responsible AI before deploying a new model. Which feature specifically refers to the model's ability to maintain consistent and appropriate behavior even when it encounters unexpected or adversarial inputs?

   • AInclusivity
   • BRobustness
   • CVeracity
   • DFairness
   Reveal Solution Discussion Select an answer first
2. 22expert · hard

   A human resources company has deployed a model to screen job applications. The compliance team has received complaints that the model is biased against applicants from certain zip codes. The data science team has run Amazon SageMaker Clarify and confirmed a bias signal. However, the team is unsure whether the bias is caused by the training data or by the model learning spurious correlations. The company wants to determine the root cause before deciding on a mitigation. Which approach should the team take?

   • AUse Amazon A2I to have human reviewers manually evaluate a sample of the model's predictions for bias
   • BRetrain the model with a more diverse dataset and compare the bias metrics before and after retraining
   • CRun a subgroup analysis with Amazon SageMaker Clarify on the training data labels and the model predictions to compare bias metrics at each stage
   • DUse Amazon SageMaker Model Monitor to track the model's performance over time and detect when the bias first appeared
   Reveal Solution Discussion Select an answer first
4. 23foundation · easy

   A data scientist is documenting the responsible AI features of a new model. Which feature is primarily concerned with ensuring the model's outputs are truthful and grounded in fact?

   • ARobustness
   • BSafety
   • CVeracity
   • DBias
   Reveal Solution Discussion Select an answer first
5. 24application · medium

   A marketing agency is using a generative AI model to create ad copy for client campaigns. A client discovers that a generated tagline closely resembles a copyrighted slogan from a competitor's campaign. The client is concerned about potential legal liability. Which legal risk of generative AI does this scenario primarily illustrate, and what is the most appropriate mitigation?

   • AEnd-user risk; implement a content moderation system to block offensive language in generated ads
   • BHallucination; implement a fact-checking pipeline to verify the originality of all generated text
   • CLoss of customer trust; implement a feedback loop to collect client complaints about generated content
   • DIntellectual property infringement; implement output filtering and human review of generated copy before publication
   Reveal Solution Discussion Select an answer first