IAPP AI Governance Practice Test

Computational bias refers to systematic errors that arise from the assumptions built into models or the data that they are trained on. In the context of artificial intelligence, this means that when training data contains biases—whether due to underrepresentation of certain groups, flawed data collection methods, or even the context in which data is gathered—those biases can propagate through the algorithms, leading to skewed results or decisions.

For instance, if an AI model is trained on historical data that reflects societal biases, the model may inadvertently learn and reproduce those biases in its predictions or outcomes. This can affect various applications, from hiring algorithms that favor certain demographics to facial recognition systems that perform poorly on specific racial or gender groups. By recognizing computational bias as a systematic error rather than an isolated incident, organizations can take proactive steps to identify, mitigate, and correct these biases, ultimately resulting in fairer and more accurate AI systems.