On a Tuesday morning in mid-September, new arrestees approached a glass-enclosed room in the basement of the Cook County Jail. One by one, they sat down on a stool, opposite a court employee at a computer on the other side of the glass. Over the next 30 minutes to an hour, their past criminal record and history of court appearances would be scrutinized in detail.
This is a computerized risk-assessment test, and in the end, each arrestee was assigned a score rating the likelihood that he or she would show up for his or her next court date, and the likelihood that he or she would commit a new offense if released on bail.
The Public Safety Assessment is the latest change to bond court procedures in Cook County. Previously, court staffers used paper forms to calculate whether a defendant was a flight risk by asking questions about the person’s residence, arrest history, employment, and drug use. Unemployed, unstably housed people with drug abuse issues were rated more of a flight risk—an evaluation experts consider to be unfairly biased against people of color and the poor. In contrast, the PSA, fully implemented in central bond court proceedings as of last spring, relies solely on documented evidence of failures to appear and past convictions.
The PSA represents a growing trend of court systems around the country using computerized risk-assessment to inform bond-setting and, in some places, even sentencing decisions. Some experts are welcoming the PSA as a tool for potentially counteracting racial and class bias in judges’ bond-setting—a bias that’s seen as contributing to the disproportionate pre-trial incarceration of poor, African-American men in the Cook County jail. But critics are concerned that bias persists even in the PSA’s supposedly neutral algorithm. And the whole debate is potentially moot in the face of a larger question: are Cook County judges even paying attention to the PSA’s scores?
The PSA—designed by the Laura and John Arnold Foundation and provided free of charge to Cook County and several other jurisdictions around the country—considers nine different variables in calculating scores for each arrestee, including prior convictions and failure to appear for court at any point in the last two years, but also age and whether the arrestee has any other pending charges at the time of the new arrest.
Scores go from one to six. The first score indicates an arrestee’s risk of committing a new crime while out on bail, and the second score indicates his or her risk of not showing up in court for the next hearing. The higher the score, the higher the risk. The PSA also issues an alert if an arrestee is considered at risk of committing a violent offense.
Based on these scores, the court staffers who administer the PSA also issue a recommendation for the sort of monitoring a defendant should have if released back into the community: an electronic ankle bracelet for house arrest, frequent home visits from a pre-trial officer, regular checks for new arrest, and/or periodic check-ins over the phone.
After all of the day’s arrestees have been processed in the jail basement, the court staffers hurry to courtroom 100, where bond court hearings begin every day at 1:30 PM. As each defendant is shuffled before a judge, the court staffers read their risk scores aloud, and recommend the conditions for release—the idea being that during the course of each typical 30- to 60-second bond hearing, the judge can consider this information if he or she chooses.
“People have been looking at Chicago bond court for a long time, and the concern has always been the length of hearing—that the judges are not really looking and seeing the defendants, most importantly their social features,” says Temple University sociologist Nicole Gonzalez Van Cleve. Her recent book, Crook County: Racism and Injustice in America’s Criminal Courts, is an expansive study of Cook County court culture based on her own experiences working in the state’s attorney’s and public defender’s offices, more than 1,000 hours of court-watching, and interviews collected with a research team from the Chicago Appleseed Fund for Justice.
Van Cleve’s findings that judges don’t take their time in considering every defendant’s case have been mirrored in court-watching reports over the years. Bond court judges are now facing a class-action lawsuit for ignoring the financial circumstances of defendants in their bail decisions, even though they are required by Illinois state law to do so.
“A lot of judges fear that if they set a fair bond and a person commits a crime while they’re out on bond, that’s gonna be used against them in an election,” says Van Cleve. “The judges are extremely resistant in many cases to integrate reforms into their practice.”
Van Cleve is hopeful that pressure on judges to pay attention to the PSA might break down the established rhythms of what she calls the “mini-fiefdoms” of each of the six bond court judges. “I don’t think there’s a great assessment tool that can be devoid of bias,” she says, “but when you have a place like Cook County where. . . [a computer assessment] takes away the power of the types of bias that I show—the type of racism and classism—it can mitigate those kinds of things.”
Although Cook County chief judge Timothy Evans, who was recently narrowly re-elected by Circuit Court judges, has hailed the PSA as an important innovation in bond court reform, he has been careful not to appear to question the authority and practices of the bond court judges. Evans’s spokesman, Pat Milhizer, is picky about the wording used to describe the goal of the PSA: it’s goal isn’t to improve bond-setting practices, he says, but to provide “one more piece of information to help [judges] make [their] decision.” The PSA “is not a substitute for a judge’s discretion,” Milhizer says.
Initial reports about the PSA’s effect on the bond decisions have been conflicting. After the PSA was piloted in central bond court in the summer of 2015, the Sheriff’s Justice Institute released a study claiming that judges ignored the court staffers’ recommendation for pre-trial release conditions in 85 percent of the 1,574 bond hearings observed between February 10 and March 29, 2016. Milhizer disputes the validity of these findings, saying that the sample was too small and that the PSA wasn’t fully implemented for judges to take into account before March 21. In a statement released Monday, Evans claims that “the use of I-bonds and electronic monitoring is climbing,” implying that the judges are taking the risk-assessment into consideration and that as a result 94 percent of defendants charged with nonviolent, non-weapons cases are now being released on recognizance bonds and house arrest.
No independent study has been conducted to track the relationship between PSA scores, release recommendations, and the judges’ bond decisions.
Skeptics of the rising use of risk-assessment algorithms are ultimately relieved that the PSA’s scores are an optional consideration. They argue that the notion of “risk” is itself highly subjective, and can contain hidden biases. The PSA’s consideration of the “likelihood that an individual will commit a new crime,” while on bond can, for example, be understood as counter to the presumption of innocence until proven guilty.
A recent investigation by ProPublica found evidence of racial bias in another algorithm used by jurisdictions from New York to California called the Correctional Offender Management Profiling for Alternative Sanctions, or COMPAS. But experts consider the PSA a much less troublesome program than COMPAS due to the simplicity of the factors it considers, and its sole focus on documented data rather than interview questions. And unlike COMPAS, the PSA doesn’t take into account arrest records, mental health, residential stability, and employment. Nevertheless, even considering a defendant’s conviction record could be problematic because discrepancies in policing between white and minority neighborhoods lead to more minorities being charged with crimes and more minorities winding up with convictions.
Given that, a risk-assessment algorithm can only predict “a system’s response to a person rather than a person’s actions,” argues Sharlyn Grace of the Chicago Appleseed Fund, a court watchdog group. “We know that arrest rates and risk assessment results will be totally different between a high school age kid in Wilmette and a high school kid in Austin.”
Traci Schlesinger, a sociologist at DePaul University, also thinks these predictive tools are overly focused on whether a defendant will miss a court date.
“[The PSA] cares way too much about failure to appear,” says Schlesinger. “They miss their court dates because sometimes they can’t get off work or get child care. People aren’t absconding; people’s lives are just a little less structured than the policymakers’ lives are.”
Instead of jailing people judged to be at risk of not showing up to court, Schlesinger and Grace both suggest phone calls and text reminders about court dates, or even sending defendants a Ventra card with a few dollars for CTA fair to help ensure they’re able to come. Even this would be cheaper than the $162 the county spends per inmate per day, keeping them in jail while they wait for their day in court.