The Role of Data in Effective AI Systems AIE Practice Questions (Page 5)

Questions 21–25

1. 21expert · hard

   A national job-matching platform is developing an AI model to recommend job postings to users. The training data is collected from user interactions, but the platform has historically had fewer users from rural areas. The team is concerned about fairness and also must comply with data protection regulations. They have limited budget for new data collection. What is the most balanced approach?

   • ARemove demographic attributes from the model to ensure fairness.
   • BCollect new data from rural areas even if it exceeds the budget.
   • CUse resampling techniques to balance the existing data and apply fairness-aware algorithms during training.
   • DDeploy the model and monitor for bias after launch.
   Reveal Solution Discussion Select an answer first
2. 22expert · hard

   An e-commerce company has a recommendation model trained on customer behavior data. The data is stored in a data lake, and the company must comply with a new regulation requiring that customer data be deleted upon request. The ML team is concerned that deleting data will degrade model performance. What is the most appropriate approach?

   • ADelete the data immediately and stop all model updates to avoid any compliance issues.
   • BImplement a data lifecycle policy that archives the data, and retrain the model periodically using only data that has not been deleted.
   • CRefuse to delete the data because it is needed for model training, and argue that the regulation does not apply to ML systems.
   • DKeep the data in the data lake but mark it as 'deleted' in the metadata to satisfy the regulation.
   Reveal Solution Discussion Select an answer first
4. 23expert · hard

   A city government is building an AI system to allocate housing assistance. The training data is historical records of past applicants, but the system's predictions are biased against certain neighborhoods. The data science team must address this while maintaining overall accuracy. Which combination of actions is most effective?

   • AUse a more complex model, such as a deep neural network, to capture subtle patterns and reduce bias automatically.
   • BRemove all sensitive attributes (e.g., neighborhood) from the dataset and retrain the model without them.
   • CAudit the data for representation gaps, apply fairness-aware preprocessing (e.g., reweighting), and evaluate the model using fairness metrics alongside accuracy.
   • DCollect more data from the underrepresented neighborhoods and retrain the model with the same algorithm.
   Reveal Solution Discussion Select an answer first
5. 24foundation · easy

   Which activity best describes data annotation in the context of AI training?

   • ACompressing data to reduce storage requirements
   • BRemoving outliers and noise from the dataset
   • CAdding descriptive labels or tags to raw data to make it usable for supervised learning
   • DEncrypting data to protect it from unauthorized access
   Reveal Solution Discussion Select an answer first
6. 25foundation · easy

   How does increasing the volume of training data generally affect AI model performance?

   • AIt can improve generalization by providing more examples for the model to learn from
   • BIt always guarantees higher accuracy
   • CIt reduces the need for data preprocessing
   • DIt eliminates the risk of bias
   Reveal Solution Discussion Select an answer first