It’s been roughly a year since I published my Bias-Variance Decomposition of Human Inequality. I hope it has been enough time for others to have thought of it. If not, then surely these high-tech companies or their foreign counterparts will have hacked my notes and found out more about it.
For companies who believe the decomposition, actions can be taken in light of new thought.
For example, the companies that used to distribute information about disparity of average compensation between equality qualified but differently raced employees, they should now supplement that data with the Bias-Variance decomposed version of the disparity.
There is a lot that I can think of in terms of using this approach to inspect and indeed modify the monitoring of equal opportunity employment. But to start we must have transparency.
The one troubling concern is this: Most of the people who can produce this decomposition are actually very well paid and respected member of our scoiety. Those who are able to understand it are also on their payroll. I mean not to disparage minorities and the excluded population. But really, I don’t think most that suffer from variance-induced inequality understand it. I mean for pete’s sake, I daren’t do more than to blog about it in an obscure blog that nobody reads, since I may be called on to seek employment again someday to provide for my family. I fear retaliation by all who benefit from the inequality, including my own future self.
The reduction of intra-group variance in an effort to reduce uncertainty means the well-off become less well-off and the less well-off become better off. Increasing correlation has similar effect of worsening the well-off’s dispositions.
Alas, I write as I believe it, that Americans should treat Americans equally; American companies should provide equal opportunity for American employees, in the highest and best sense that we understand.
May the root be with us! May root save us all!