Bias detection and mitigation

The trajectory of the word “bias” is interesting in that, in the last 15 years, it’s come full circle. Originally, bias was arguably a statistical term. Formally, it implied that a sample size was improperly constructed, giving excessive weight to certain variables. Statisticians developed numerous methods to identify and reduce bias to evaluate studies properly, such as those used in randomized control trials in public health or policy evaluations in econometrics. Basic tactics include making sure that the treatment and control groups are roughly the same size and have roughly the same characteristics. Without a guarantee of that basic mathematical equivalence, or more realistically as close to it as the research team can get, it’s difficult to trust that the results of a study are truly valid. The results themselves are subject to bias, simply indicating the presence or absence of basic characteristics, rather than implying...