What is historical bias?

• A. Bias caused when data reflects existing societal inequities
• B. Bias caused by temporal drift in data
• C. Bias caused by incorrect labeling
• D. Bias due to missing values

Answer: (A)

Historical bias happens when the data used to train a model reflect past social inequalities. Because the model learns from those historical outcomes, it can reproduce or even amplify unfair patterns—disadvantaging or privileging certain groups based on race, gender, socioeconomic status, and other factors that were unequally treated in the past. This kind of bias is about what the data actually encode: discriminatory practices, unequal opportunities, and biased decisions that existed historically.

This is different from temporal drift, which is about how data distributions change over time and may require updating the model to new patterns. It also isn’t about labeling mistakes or missing values—those are data quality issues, not reflections of systemic inequities captured in the data.