I hear two very different claims about this. One side says IQ tests are obviously biased because different groups score differently on average. The other side says that score differences are not the same as bias and that the tests actually predict outcomes equally well across groups. I want to understand what psychometricians actually mean by bias and whether the evidence supports the claim that modern IQ tests are biased against minority groups.
This is one of the most important distinctions in all of psychometrics and it gets collapsed constantly in popular discussion. In everyday language people use bias to mean unfair or disadvantaging. In psychometrics bias has a specific technical meaning: a test is biased if it measures something different across groups, or if it predicts outcomes less accurately for one group than another. By that definition, the research is actually pretty consistent. Modern well designed cognitive assessments show measurement equivalence across racial and ethnic groups, meaning the factor structure, item functioning, and predictive validity hold up across groups. The tests predict academic performance, job performance, and other outcomes at roughly the same rate for Black, white, and Hispanic test takers.
What does exist are mean score differences between groups. But a mean difference is not evidence of bias by itself. A thermometer is not biased just because people in different climates have different average temperatures. The question is whether the instrument is measuring the same thing accurately across groups, and the evidence says it largely is.
Score gaps and test bias are genuinely different claims and conflating them does not help anyone. A biased test measures differently across groups. A test that accurately measures real differences in developed abilities is doing its job, even if those results are uncomfortable.