The Role of Data in Effective AI Systems AIE Practice Questions (Page 7)

Questions 31–35

1. 31foundation · easy

   Which strategy is most effective for mitigating bias in an AI system?

   • AEnsuring the training data is representative and includes diverse examples from all relevant groups
   • BIncreasing the model's complexity to capture more patterns
   • CRemoving all demographic attributes from the dataset
   • DUsing only the most recent data available
   Reveal Solution Discussion Select an answer first
2. 32foundation · easy

   Which data quality dimension is most directly concerned with whether data values reflect the real-world entities they represent?

   • ACompleteness
   • BAccuracy
   • CConsistency
   • DTimeliness
   Reveal Solution Discussion Select an answer first
4. 33expert · hard

   A recruiting firm is building an AI tool to screen job applicants. The training data is resumes from the past five years, and the model learns to favor candidates from certain universities. The firm wants to reduce this bias without significantly reducing the number of qualified candidates. Which approach is most appropriate?

   • ABlind the model to all demographic information and assume the bias will disappear.
   • BAnalyze the training data for proxy variables (e.g., university prestige correlated with demographic attributes) and apply bias mitigation techniques such as adversarial debiasing.
   • CUse a larger dataset of resumes from more universities to naturally reduce the bias.
   • DRemove the university name from the resume text and retrain the model on the remaining features.
   Reveal Solution Discussion Select an answer first
5. 34expert · hard

   A global e-commerce company is building a product recommendation model. They have data from multiple countries, but the data volume is highly skewed: 80% of the data comes from the US, and the remaining 20% comes from other countries. The model performs well for US users but poorly for users in other countries. What is the most effective strategy to improve performance for non-US users?

   • ACombine all data into one dataset and train a single global model, assuming more data is always better.
   • BTrain separate models for each country using only that country's data, ignoring the US data.
   • CRemove the US data and train only on the non-US data to ensure the model is not biased toward the US.
   • DUse domain adaptation or transfer learning to leverage the US data while fine-tuning on the smaller non-US datasets.
   Reveal Solution Discussion Select an answer first
6. 35foundation · easy

   What is the primary purpose of data cleaning in the preprocessing phase?

   • ATo handle missing values, remove duplicates, and correct errors
   • BTo split data into training and testing sets
   • CTo reduce the number of features in the dataset
   • DTo convert categorical data into numerical values
   Reveal Solution Discussion Select an answer first