As an aside, b_b, the link you sent me is a good example of the importance of the lesson that "you must not fool yourself—and you are the easiest person to fool." I didn't read the article until now, and I see it both amplifies the conclusions of the study and diminishes the reservations. The researchers have "demonstrated conclusively" that "racial bias affects judicial sentencing;" they "offer statistical proof" that judges vary in their treatment of race. When the researchers say "We find evidence ... providing support for the model ..." the copywriter translates this to "race is affecting sentencing decisions" in the same paragraph. Most egregiously, "the technique eliminates the problem of 'unobserved variables,'" which the researchers took pains to point out as a problem with their study. Many people will read the press release and not look closely at the study. It's always a good idea to cut out the middleman!
Interestingly, the data only say that racial bias exists, but not whom it is working against (or for). However, we do have evidence that black people were targeted for sentencing disparity in mandatory minimum sentencing laws. The Supreme Court has ruled against these laws (specifically harsher sentences for crack than powder cocaine), but I don't think we're completely rid of them.