Bias CT-AI Practice Questions (Page 3)

Questions 11–15

1. 11foundation · easy

   In the context of AI-based systems, which statement best defines bias?

   • AA systematic error that leads to unfair outcomes or decisions, often reflecting societal prejudices or flawed data
   • BThe difference between training accuracy and test accuracy
   • CAny deviation from a perfect prediction, including random noise in the data
   • DThe variance in model predictions when trained on different random samples
   Reveal Solution Discussion Select an answer first
2. 12expert · hard

   A healthcare provider is deploying an AI system to recommend treatment plans for patients with chronic conditions. The system was trained on a diverse dataset and performs equally well across demographic groups. However, during a pilot deployment, clinicians notice that the system's recommendations are often overridden for patients from a particular ethnic group. The clinicians explain that they do not trust the system's recommendations for these patients because the system sometimes suggests treatments that conflict with the patients' cultural preferences. Which impact of bias is most directly demonstrated?

   • AIncreased computational cost due to the overrides
   • BIncreased legal liability for the healthcare provider
   • CReduced model accuracy for the ethnic group
   • DReduced user trust, leading to the system's recommendations being overridden
   Reveal Solution Discussion Select an answer first
4. 13foundation · easy

   Which of the following is a source of bias in AI-based systems?

   • AThe color scheme used in the user interface of the AI application
   • BData that is not representative of the target population
   • CThe choice of programming language used to implement the model
   • DThe speed of the model's inference at runtime
   Reveal Solution Discussion Select an answer first
5. 14foundation · easy

   Which technique is commonly used to mitigate bias during data collection?

   • AReducing the number of features
   • BIncreasing the model's complexity
   • CUsing a different programming language
   • DEnsuring the dataset is representative of the target population
   Reveal Solution Discussion Select an answer first
6. 15application · medium

   A social media platform deploys an AI system to automatically moderate hate speech. The system was trained on a dataset of comments that were labeled by a team of human moderators. The moderators were given vague guidelines and often disagreed on what constituted hate speech. The deployed system now frequently flags comments from minority groups as hate speech, even when they are not. Which type of bias is most directly demonstrated?

   • APrejudice bias, because the moderators intentionally targeted minority groups
   • BSample bias, because the training data did not include enough comments from minority groups
   • CExclusion bias, because minority groups were excluded from the training data
   • DMeasurement bias, because the labels used to train the model were inconsistent and unreliable
   Reveal Solution Discussion Select an answer first