Testing for Algorithmic, Sample and Inappropriate Bias CT-AI Practice Questions (Page 3)

Questions 11–15

1. 11application · medium

   A company is building a chatbot for customer service. The training data was collected from chat logs of customers who contacted support during business hours. The chatbot performs well during the day but poorly at night. The team suspects that night-time customers have different issues. Which type of bias is this, and what testing approach would help?

   • ASample bias; test the chatbot with a dataset that includes night-time chat logs.
   • BInappropriate bias; conduct a fairness audit on response times across different customer segments.
   • CAlgorithmic bias; adjust the chatbot's natural language processing parameters.
   • DSample bias; collect more data from business hours to improve the model.
   Reveal Solution Discussion Select an answer first
2. 12foundation · easy

   Which statement best defines algorithmic bias in an AI system?

   • AA type of bias that only occurs when the training data is not representative of the target population.
   • BA deliberate attempt by developers to encode their personal preferences into the model's decision logic.
   • CA systematic and repeatable error in a computer system that creates unfair outcomes, such as privileging one arbitrary group of users over others.
   • DA random error in the training data that causes the model to make occasional mistakes on new inputs.
   Reveal Solution Discussion Select an answer first
4. 13foundation · easy

   What is a potential impact of sample bias on an AI system's quality?

   • AThe model will always be more accurate on the majority group but never on minority groups.
   • BThe model will have no impact on stakeholders because bias only affects the model's internal parameters.
   • CThe model may perform poorly on underrepresented groups, leading to reduced overall accuracy and fairness.
   • DThe model's predictions will be completely random for all users.
   Reveal Solution Discussion Select an answer first
5. 14application · medium

   A government agency uses an AI system to flag fraudulent benefit claims. The system is found to have a higher false-positive rate for applicants from certain immigrant communities. The agency is concerned about the impact on these stakeholders. What is the most appropriate testing technique to assess the impact of this bias?

   • AReview the model's decision tree to identify which features are used.
   • BMeasure the false-positive rate separately for each demographic group and compare them.
   • CRetrain the model with a balanced dataset across all demographic groups.
   • DConduct a survey to ask affected applicants how they feel about the system.
   Reveal Solution Discussion Select an answer first
6. 15foundation · easy

   Which of the following is a common source of sample bias during the data collection phase?

   • AChoosing a machine learning algorithm that is too simple to capture complex patterns.
   • BCollecting data only from users who have access to a specific smartphone app, excluding non-users.
   • CSetting the model's fairness threshold too high, causing many false positives.
   • DUsing an optimization function that penalizes errors on minority groups more heavily.
   Reveal Solution Discussion Select an answer first