ABSTRACT
Bias is a systematic error. Test bias refers to a systematic error (in measurement, prediction, etc.) as a function of group membership. The two broad categories of test bias include predictive bias and test structure bias. Predictive bias refers to differences between groups in the relation between the test and criterion—i.e., different intercepts and/or slopes of the regression line between the test and the criterion. Test structure bias refers to differences in the internal test characteristics across groups, and can be evaluated with approaches such as measurement invariance and differential item function. When detecting test bias, it is important to address it, and there are a number of ways of correcting for bias, including score adjustment and other approaches that do not involve score adjustment. However, even if a test does not show bias does not mean that the test is fair. Fairness is not a scientific question but rather a moral, societal, and ethical question; there are many different ways of operationalizing fairness. Fairness is a complex question, so do the best you can and try to minimize negative impact.
