Yes. AI bias is not just about protected classes. It is also about how a model determines an object, such as the definition of a guard rail for an autonomous vehicle. Following are three common types of bias an auditor/regulator may be looking for:
• Latent Bias
An algorithm may incorrectly identify something based on historical data or as a result of an existing stereotype. This type of bias is commonly found in financial, healthcare, and automotive industries.
• Selection Bias
Selection bias occurs when a data set for training AI models is not accurate and overrepresents one particular group and underrepresents another.
• Emergent Bias
Emergent bias results from using and relying on algorithms across new or unanticipated contexts. In other words, the algorithm was designed for one purpose but is used for another.