Testing for bias · Question 46 of 60

A team wants to find out whether its model produces biased results, but does not know whether the cause lies in the data or in model development. Which approach detects bias regardless of its cause?

1 answer is correct. Hard

1. A Only static analysis of the model code, since bias always arises in the algorithm
2. B Dynamic testing with a known unbiased, representative dataset, evaluated per sensitive group
3. C Only a review of the dataset documentation, since bias basically always stems from the training data
4. D Only a review of the data preparation, because all causes become visible there