I don't know if you're doing this (again and again) intentionally or not, but racial bias is not statistical bias[1]!
> It's silly to say that by focusing on a technology project and ignoring the people involved, we alienate people.
How do you know it's silly? Have you done any research on that? Because people who have actually done some research have come to the conclusion that it is your assumption that is silly[2].
> "meritocracy" is one of our core values
Meritocracy is not a value (or not a distinguishing value). It's what every (long surviving) system believes itself to be. The term is ironic and generally means "a system that deludes itself about its own fairness". The term was coined as a satirical joke, and has no other well-defined meaning.
[1]: Unless your statistics includes every possibly pertinent variable, or you manage to prove total independence of your model's variables from all other variables.
White like Satya Nadella and Sundar Pichai, you mean?
Google is 60% white. Facebook is 55%. Microsoft is 59%. USA is 63%. Do you have any evidence that HN is whiter than Google and Facebook? Why not say what you really mean, namely that there are too many Asians?
Or machine learning algorithms that courts use to predict recidivism rates in determining sentencing have much higher false positive rates for African-Americans than whites.[4]
Not according to the R script written by the very authors of the piece you cite. That script is unable to find statistically significant evidence of bias. They found very strong evidence (p < 0.001) that the statistical model is highly effective, albeit imperfectly calibrated [1].
https://www.chrisstucchio.com/blog/2016/propublica_is_lying....
It's silly to say that by focusing on a technology project and ignoring the people involved, we alienate people. Where I work we have similar views - "meritocracy" is one of our core values, not the dirty word it's become in the valley. As the only member of my race at this company this makes me feel welcome - I'm as valuable as the things I build, who cares about my race?
[1] This means that higher score means higher recidivism rate, but a 60% score doesn't necessarily mean 60% recidivism probability. Probably should have run isotonic regression or something like that. https://en.wikipedia.org/wiki/Isotonic_regression