The case for risk assessments despite allegations they are racist

Here's an unpopular take that I'll probably regret offering:

Grits finds himself a tad annoyed with well-intentioned liberals crowing from the rooftops that risk assessments in criminal justice fields are "racist," and wishes these debates allowed for more nuance.

While I agree that criminal-justice data skews to overstate risk for black folks, I don't necessarily agree that foregoing data-driven risk assessments achieves better outcomes. Judges making seat-of-the-pants decisions rely on the same discriminatory data, and incarcerate people of all races at greater rates than risk-assessments would dictate.

Let's cut to the chase. Consider a thought experiment in which a jurisdiction is 85% white and 15% black. In this jurisdiction, based on judges' stand-alone decisions, 100 people are being held pretrial, and the proportions by race are 60 white folks and 40 black folks, so there's a significant racial disparity.

Now lets say the risk-assessment algorithm also discriminates because of the use of data from a system embedded with a racist, even white-supremacist ideology, as well as historical disparities in violent crime rates by race.*

For the sake of argument, let's say the risk assessment labels 1 in 3 white people incarcerated as "low risk" and recommends release, but only 1 in 4 black defendants. Now let's assume judges release every inmate labeled "low risk." What would be the result?

Under this scenario, 40 white people would be incarcerated compared to 30 black people. Thus, the percentage of black folks incarcerated would go up, from 40 to 43 percent. So, it's possible to accurately say that use of the risk assessment INCREASES racial disparities.

But which outcome is better from a civil-rights perspective - 40 black folks locked up in jail or 30?

I'm a middle-aged white guy and certainly can't and won't attempt to speak for black people. But my own belief is that the outcome under the risk-assessment model is more tolerable. Others' mileage may vary.

What am I missing here?

CAVEAT: There are many offenses where no risk assessment need be used at all. In the Harris County bail settlement, 85% of misdemeanor defendants will be released without a bail determination hearing of any sort. Where that's not possible, IMO risk assessments are superior to judges making the determination on their own, at least until someone convinces me otherwise.

*Violent crime rates among black folks have been declining dramatically in recent years, falling at greater rates, even, than overall crime declines. But the rates remain higher than other racial categories. According to the 2017 National Crime Victimization Survey, "Based on victims reports, there were about four fifths as many white [violent-crime] offenders as the percentage of whites in the population, [and] about twice as many black offenders as the percentage of blacks in the population." Because most victims of violent crime are attacked by someone of the same race, this also translates into more black victims: "For the year 2015, blacks represented 13 percent of the nation’s population, yet accounted for 51 percent of all homicide victims." This dynamic is a source of some but by no means all of the disparities attributed to risk assessments.