What Kind of Fairness Is Fair? A Comment on Conway and Roberts
评论康威和罗伯茨关于直接回归与反向回归对应两种公平概念的观点，指出其结论需要商榷，对研究就业歧视的学者有参考价值。
In his previous work, Roberts (1979,1980) emphasized problems of measurement and possible bias involved in using direct regression, where reward is the dependent variable, and the independent variables consist of a vector of qualifications. He also suggested that these problems could, to a large extent, be solved by using reverse regression, where reward is the independent variable, and the dependent variable consists of a measure of qualifications. The debate on these issues is continuing. (See Blattenberger and Michelson 1980, Finkelstein 1980, Madden 1982, and Blau and Kahn 1983.) In Reverse Regression, Fairness, and Employment Discrimination, however, Conway and Roberts (1983) put primary emphasis on the fact that there are two distinct concepts of fairness associated with direct and reverse regression and that these different perspectives are basic to understanding discrimination. They also point out, quite correctly, that these two concepts will only occasionally be in agreement. Up to this point, their exposition of these issues is clear and helpful. It is when they go on to claim that both concepts are needed to evaluate fairness, and that contradictions between the two pose serious problems for lawyers, statisticians, and judges alike that their conclusions need to be challenged.