When auditing a machine learning pipeline, which activity in the data preprocessing stage poses the HIGHEST risk of introducing systematic bias into the model?
Select an answer to reveal the explanation.
Short Explanation and Infographic
Here's the deal — b is correct because imputing missing values using statistics from a majority group systematically underrepresents minority group characteristics, embedding structural bias into the training data that propagates through model training. A (normalization) and D (one-hot encoding) are standard preprocessing steps that do not inherently introduce bias.
Full explanation below image
Full Explanation
B is correct because imputing missing values using statistics from a majority group systematically underrepresents minority group characteristics, embedding structural bias into the training data that propagates through model training. A (normalization) and D (one-hot encoding) are standard preprocessing steps that do not inherently introduce bias. C (deduplication) generally improves data quality.