Bias, Testing, and Assessment

Bias in testing and assessment refers to the extent to which a test or other formal sample of performance is constructed, administered, scored, or summarized so that identifiable subgroups of examinees are systematically advantaged or disadvantaged. Such inadvertent positive or negative impact on subgroups is independent of the content and skills to be assessed on the examination. Samuel Messick called this kind of bias “construct-irrelevant variance” and described the systematic and predictable error that can occur with test results. This entry addresses what may cause bias and how it may be avoided, what techniques have been used and their value, and the greater problem of ensuring fairness in the operation of an educational system, where assessment and inferences from results may be a major component.