Testing for Algorithmic, Sample and Inappropriate Bias CT-AI Practice Questions (Page 6)

Questions 26–28

1. 26foundation · easy

   Which scenario best illustrates sample bias in an AI system?

   • AA model's performance degrades over time because the real-world data distribution shifts after deployment.
   • BA chatbot gives offensive responses because its training data contains biased language from online forums.
   • CA loan approval model uses a feature that is a proxy for race, leading to discriminatory decisions.
   • DA facial recognition system is trained only on photos of people with light skin tones and performs poorly on darker skin tones.
   Reveal Solution Discussion Select an answer first
2. 27foundation · easy

   Which testing technique is most appropriate for detecting sample bias in a training dataset?

   • AAnalyzing the distribution of protected attributes in the training data compared to the target population.
   • BReviewing the model's source code for hardcoded rules.
   • CPerforming a stress test with extreme input values.
   • DMeasuring the model's accuracy on a holdout test set that is also biased.
   Reveal Solution Discussion Select an answer first
3. 28foundation · easy

   Which of the following is a common source of algorithmic bias during the model development phase?

   • ASelecting features that are proxies for protected attributes, such as using zip code as a proxy for race.
   • BUsing a dataset that contains duplicate records from a single source.
   • CLabeling errors in the training data that misclassify certain groups.
   • DCollecting data only from a specific geographic region.
   Reveal Solution Discussion Select an answer first