Wednesday, May 22, 2002
In other reading. I'm reading more than the Wolfram book right now. One other read is Political Numeracy, by Michael Meyerson. It's about the intersection between mathematics and law, pretty much in the USA. He discusses matters such as how the Electoral College concentrates voting power to the individual (by having multiple smaller elections instead of one great big one), the Three-Fifths Compromise which led to the adoption of the US Constitution, and supermajorities for overriding presidential vetoes and approving treaties.

I just finished Chapter 5, which is an analysis of affirmative action. Meyerson's analysis is insightful. He defines "good faith" participants in the discussion to mean (1) that they're honest and open to persuasion, (2) that they're nonracist and don't believe in some intrinsic superiority of any racial group, and (3) that they're nonjingoistic and that they don't believe that members of their own racial group ought to benefit at the expense of members of some other group. (Beyond some introductory examples, he sticks largely to racial matters, but his analysis is extendable to any cohort identifiable by status or behavior.)

He then goes through a nice description of what someone of his definition of good faith ought to desire from employment policy: that the fraction of qualified people who are successful at getting jobs increases, while the fraction of unqualified people who are successful at getting jobs decreases. The he gets to the nub issue: What kinds of measures are used in decision making about job applicants, and how good are those measures at predicting eventual success in the job.

He ends up with the following 10 "points of agreement" regarding affirmative action policy that he believes people of (his definition of) good faith ought to be able to adopt as a framework for discussion in trying to determine the appropriate policy. Quoting, his points are:

  1. It is possible that a person who disagrees with me is acting in good faith.
  2. The selection system I support represents an imperfect model. It does not create a pure meritocracy but uses various criteria to approximate merit.
  3. Due to any system's imperfections, some qualified candidates, African American and white, are bound to be unsuccessful; moreover, some unqualified applicants will prevail regardless of the system used.
  4. Not all numerical disparities indicate discrimination. But significant numerical disparities, coupled with our understanding of the history and current existence of discrimination, may indicate a problem.
  5. Quantifiable factors such as test scores are imperfect measures for the important subjective and unquantifiable attributes, such as most qualified.
  6. Race is an imperfect model for identifying those who are disadvantaged or victims of discrimination.
  7. The weaknesses of any modeling system, even one created in good faith, create innocent victims.
  8. A race-neutral mechanism does not guarantee that qualified applicants of all races will be identified. A race-based mechanism does not guarantee that only qualified applicants will be benefited.
  9. Increasing our sophistication in defining qualifed in an ideal sense and improving the proxies we use to approximate it are essential for avoiding the problem of the innocent victim.
  10. Good faith will also be revealed by the energy people are willing to expend on finding acceptable solutions.

As Meyerson says, "a recognition that all systems are imperfect will, at least, remove some of the self-righteousness from the debate. Moreover, acknowledgement that the problem of innocent victims affects both sides may encourage a more concentrated effort to find qualified people whoever they may be."