Which concept describes bias in training data that can be learned and perpetuated by models?

Get ready for the GARP Risk and AI Exam with flashcards and multiple choice questions. Each question comes with hints and explanations. Prepare for success!

Multiple Choice

Which concept describes bias in training data that can be learned and perpetuated by models?

Bias in training data that a model can learn and perpetuate is called hidden bias. It happens when the data used to train the model contains skewed patterns, historical prejudices, or labeling inconsistencies. Since the model learns from these patterns, the bias becomes encoded in the model’s parameters and can show up in its predictions on new data, often reproducing the same unfair or biased outcomes. This concept is distinct from privacy threats, which focus on protecting personal data; manipulation, which involves altering data or results; and lack of accountability, which concerns governance and responsibility. Hidden bias explains why models can carry forward biases from the data they were trained on.

Bias in training data that a model can learn and perpetuate is called hidden bias. It happens when the data used to train the model contains skewed patterns, historical prejudices, or labeling inconsistencies. Since the model learns from these patterns, the bias becomes encoded in the model’s parameters and can show up in its predictions on new data, often reproducing the same unfair or biased outcomes. This concept is distinct from privacy threats, which focus on protecting personal data; manipulation, which involves altering data or results; and lack of accountability, which concerns governance and responsibility. Hidden bias explains why models can carry forward biases from the data they were trained on.

More practice questions