One of the great injustices in today's justice system is that so many people are punished long before they have been found guilty.
They are punished by being kept in pretrial detention for months or even years, either because they were denied bail or because they could not afford it. They routinely lose their jobs, housing and even custody of their children. Many, even innocent people, feel immense pressure to plead guilty because it will get them out of jail.
The ostensible reasons these people are kept in pretrial detention are that they either pose a significant risk to the community or are unlikely to appear for trial. These risks have traditionally been determined by judges. In an effort to make the process more scientific, however, many jurisdictions have put computer-generated pretrial risk assessments in place to help judges decide.
These pretrial risk assessment algorithms take the available data on an individual's present criminal charge, demographics and criminal history to recommend whether the person should be released before trial.
As we've discussed in the past, however, studies have shown that many, if not all, of these algorithms merely reinforce existing racial and socioeconomic biases, generating assessments that are much more favorable for wealthy whites than for others.
Keep in mind that everyone held in pretrial detention is legally presumed innocent. Yet pretrial detainees make up a huge percentage of those incarcerated in the U.S.
"I think people discount the well-being and liberty of people accused of crime, because we see them as already guilty," says a professor at the University of Georgia School of Law. "I think we've been willing to tolerate exorbitant rates of pretrial detention in part because of that."
In Chicago, risk assessments did change judges' behavior
In Cook County, Illinois, judges have been using the Public Safety Assessment (PSA) tool from Arnold Ventures. Reformers hoped the tool would reduce the percentage of people held in pretrial detention, which was then 90% of all defendants.
It did, among those deemed low-risk. That said, it created a correspondingly sharp increase in pretrial detention among those the tool deemed riskier.
Yet research shows that the overall risk that someone will reoffend is about 1% to 3%, at most. The risk that they won't return for trial is equally small, especially when courts take simple steps like texting defendants with reminders.
Perhaps, rather than attempting to predict which defendants pose risks, we should instead consider how low the risks actually are. Unless there is a clear threat or flight risk, pretrial release with no or low bail should be the default option for our courts.