Exams / AIF-C01 / Bias, variance and datasets / Question 2
AIF-C01 practice question: Bias, variance and datasets (2 of 2)
Guidelines for Responsible AI A free question from our original AIF-C01 practice exam. Try it, then open the answer to see why each option is right or wrong.
A lender is collecting data to train a credit risk model and wants the model to treat applicants fairly. Which TWO dataset characteristics are MOST important? (Choose two.)
Choose two answers.
- AA. The data is diverse and representative of all applicant groups
- BB. The data comes from a single branch to keep it consistent
- CC. The labels are accurate and checked for historical bias
- DD. The dataset is as large as possible, whatever its source
- EE. Applicants with incomplete records are silently removed
Show the answer and explanation
Answers: A, C
- A. Right: Correct. If some groups are under-represented, the model may perform worse for them.
- B. Wrong: One branch is unlikely to represent the full applicant population, which can introduce bias.
- C. Right: Correct. Labels that encode past unfair decisions teach the model to repeat them.
- D. Wrong: Volume does not fix bias. A large but skewed dataset still produces a biased model.
- E. Wrong: Dropping records without analysis can remove certain groups and skew the data.
Explanation
Fair models start with inclusive, diverse, balanced and accurately labelled data. Teams should also analyse model performance for each subgroup, audit labels for bias, and involve human reviewers, because a model can be accurate overall but unfair to a particular group.
Study
Keep practising
Answer questions with a score at the end, or practise just this topic with the explanation after each one. No sign-up needed.