Bias CT-AI Practice Questions (Page 2)

Questions 6–10

1. 6foundation · easy

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

   • AThe system may produce unfair or discriminatory results for certain groups
   • BThe system will become slower over time
   • CThe system will always have lower accuracy
   • DThe system will require more computational resources
   Reveal Solution Discussion Select an answer first
2. 7foundation · easy

   How can bias in an AI system affect user trust?

   • AUsers will trust the system more because it is automated
   • BTrust is only affected by system accuracy, not bias
   • CUsers will not notice bias unless it is explicitly reported
   • DUsers may lose confidence in the system if they perceive unfair treatment
   Reveal Solution Discussion Select an answer first
4. 8application · medium

   A bank is developing an AI system to approve small business loans. The training data was collected from the bank's historical loan applications over the past 15 years. The data shows that loan applications from female-owned businesses were rejected at a higher rate than those from male-owned businesses during the first 10 years, but this disparity disappeared in the last 5 years. The model trained on the full dataset still shows a higher rejection rate for female-owned businesses. Which type of bias is most directly demonstrated by this scenario?

   • AExclusion bias, because female-owned businesses were excluded from the training data
   • BPrejudice bias, because the bank's loan officers intentionally discriminated against female applicants
   • CMeasurement bias, because the loan approval criteria were not consistently applied over time
   • DSample bias, because the historical data does not represent the current applicant population
   Reveal Solution Discussion Select an answer first
5. 9application · medium

   A city government is using an AI system to prioritize which buildings should be inspected for fire code violations. The system was trained on data from past inspections. The data was collected by inspectors who, due to staffing constraints, focused their inspections on neighborhoods with older buildings. The model now assigns lower priority to newer buildings, even those with known safety issues. Which type of bias is most directly demonstrated?

   • AMeasurement bias, because the inspection results were not recorded consistently
   • BPrejudice bias, because the inspectors intentionally avoided newer buildings
   • CSample bias, because the training data over-represents older buildings and under-represents newer ones
   • DExclusion bias, because newer buildings were completely absent from the training data
   Reveal Solution Discussion Select an answer first
6. 10application · medium

   An insurance company is building an AI system to predict the risk of car accidents. The training data includes a feature for 'postal code'. The model learns that certain postal codes are associated with higher risk. However, these postal codes also correlate strongly with neighborhoods that have a high proportion of minority residents. Which type of bias is most directly demonstrated?

   • AMeasurement bias, because the postal code feature is not accurately recorded
   • BPrejudice bias, because the model is using postal code as a proxy for race
   • CExclusion bias, because minority residents were excluded from the training data
   • DSample bias, because the data was not collected from all postal codes
   Reveal Solution Discussion Select an answer first