Which of the following is NOT a common bias type in AI?

• A. Representation bias
• B. Measurement bias
• C. Data drift bias
• D. Historical bias

Answer: (C)

Understanding biases in AI starts with recognizing where unfair outcomes come from. Data drift describes a change in the data distribution over time, which can cause a model to become less accurate as contexts shift. It’s about evolving inputs, not a systematic tendency baked into how data are collected or labeled. Because of that, it isn’t categorized as a common bias type. The idea behind the other terms is different: representation bias occurs when the training data don’t reflect the real population, leading to skewed performance; measurement bias happens when data collection or measurement tools introduce consistent errors; historical bias arises when past decisions or societal prejudices are embedded in the data, biasing outcomes. So, data drift stands apart from these predefined bias types, making it the correct choice for NOT being a common bias type.