Testing for bias · Question 24 of 60
Which test approaches are suitable for detecting bias?
4 answers are correct. Medium
Show answer
Answer
- A Static analysis of the data preparation programs and the model code
- B Reviews of the dataset documentation
- D Exploratory data analysis of the training data
- E Reviews of the ML workflow, especially data preparation
Explanation
Suitable are reviews of the ML workflow and the dataset documentation, static analysis of the data preparation programs and the model code, exploratory data analysis, dynamic testing with a known unbiased dataset, label correctness testing and disparate impact analysis. Load testing and neuron coverage serve other purposes.