In Kansas, Algorithms Might Rewrite Who Stays In Jail And Who Bails Out | KCUR

In Kansas, Algorithms Might Rewrite Who Stays In Jail And Who Bails Out

Jan 22, 2019

Let’s say you’re arrested. You’re booked into your local jail and the district attorney decides to press charges.

The next day, you make your first court appearance in front of a judge, who then has to make a decision. Let you go home before trial — or keep you in jail?  And under what conditions?

A pretrial stint in jail could last for weeks and keep you from your children, your car payments and your job. You could lose access to resources you need to win your case. You could spend sleepless nights in a crowded, dangerous, poorly maintained detention center — all without being convicted of a crime.

Unless, of course, the judge lets you bail out: pay a certain amount of cash, get released from jail, and get the money back if you show up for court.

But the amount might be too high for you to afford. So you might pay a bond company a nonrefundable percentage of your total bail. In return, that company bails you out — and hunts you down for the full amount if you skip your court date. Can’t rustle up the money for a bond company? Sit things out in jail.

Cash bail is supposed to ensure that you come back for trial. But the traditional bail system draws fire for discriminating against the poor and disproportionately affecting the livelihoods of racial minorities, who are already more likely to be arrested and charged with crimes.

Now Kansas courts want to rethink the concept. The Kansas Supreme Court has convened a task force of judges, attorneys and corrections officials to study pretrial reforms and submit recommendations in mid-2020. The state could eventually join California, New Jersey and others by overhauling its conditions for awarding bail in state courts.

“We need to find out what works best in our state,” said Judge Karen Arnold-Burger, chair of the Pretrial Justice Task Force and chief judge of the Kansas Court of Appeals. “Before we do something that's going to completely disrupt someone's life, and before they’ve been convicted of any crime, we want to make sure we know how we're making the most educated decision.”

An ancient system

Bail originated in England centuries ago.

Both the U.S. and the Kansas Constitutions guarantee the right to bail in non-capital offenses — crimes not punishable by the death penalty.

Proponents of cash bail argue that it’s essential for keeping people out of jail and ensuring they still show up to their trial because bondsmen are inherently motivated to get their money back.

But tides have shifted in recent decades. Advocates who oppose bail say it discriminates against poor people by basing release decisions on a person’s ability to pay. Some say any form of pretrial detention violates the U.S. Constitution by imprisoning a person before they’re convicted.

The justice system and law enforcement also recognize that jail time can wreak havoc on the lives of people whose unproven charges are often low-level offenses that cause little or no harm to others. The effects are harder on the poor.

“There are lots of collateral consequences to periods of detention, even short periods of detention, from loss of jobs to loss of schooling, to loss of benefits, public assistance, veteran’s benefits,” said Arnold-Burger. “The list is long.”

A new way of assigning bail

One possible alternative is a statistics-based system already used by some of the most populous counties in Kansas. Data-based pretrial risk assessment takes information gathered from years of arrests and trials. It pinpoints which characteristics are related to two outcomes the corrections system wants to avoid: a defendant skipping their court date or committing another crime once they’re released on bail.

Depending on which factors they meet, defendants are assigned a risk level with a corresponding recommendation for a bail amount. In Johnson County, which has used a form of risk assessment since 2009 for minor crimes, that amount ranges from $0 to more than $50,000.

The county has had a custom risk assessment based on its own arrest data since 2014 and has revised it multiple times. County corrections director Robert Sullivan said the system is currently being used to assess people who have been charged with crimes that are likely to carry a sentence of probation, rather than prison.  The county plans to expand the assessment to all detainees in February.

“What we’re trying to do is move away from basing release from jail on (a defendant’s) ability to make bond,” said Sullivan, “and base it more on a detainee’s risk of failing to appear for court.”

Johnson County corrections director Robert Sullivan.
Credit Nomin Ujiyediin / Kansas News Service

The county hired University of Missouri-Kansas City professor Alex Holsinger to develop the system. Holsinger analyzed data from 2011 and 2012 to find which characteristics were most related to a failure to appear in court or committing another crime.

Holsinger found that people who live outside of the state are more likely to skip court than people who live in Kansas. Prior arrests, substance abuse and being unemployed are other risk factors.

The nature of the charge also has an effect.

“If your charge is DUI related, that’s considered a risk factor. If it’s drug related, that’s considered to be even more of a risk factor,” Holsinger said. “Somebody who doesn’t have any (previous) jail time is considered less risky than somebody who does.”

The factors are tallied into a numerical score. That’s handed to judges, who decide whether to set cash bail or other conditions of release based on a defendant’s risk level and other factors.

The result is a system that recommends the highest cash bonds — of $50,000 or more — for people who are deemed to be “extreme” risk. That category includes those who are charged with felony crimes that are likely to carry a prison sentence if the defendant is found guilty.

Much lower amounts — between $0 and $2,500 — are recommended for people deemed “low,” “moderate” or “significant” risk. Those defendants are generally charged with misdemeanors, or felonies that carry a presumptive sentence of probation, rather than prison.

For example, a person who is charged with a misdemeanor DUI, who lives outside of Kansas but doesn’t have any other risk factors, would receive a score of 3. That means the person is considered “low risk,” and could be released, not on cash bail, but on a personal recognizance bond — a signature promising to come back.

But another person charged with the same crime might receive a higher score if they are unemployed, have been in jail before, have a history of substance abuse, live outside of Kansas, and were first charged with a crime below the age of 21 — all risk factors.  That person would receive a risk score of 8 and would be considered “serious risk,” which carries a recommended bond amount of up to $2,500.

Holsinger said the scoring system helps judges make more objective decisions while still allowing them discretion.

“The risk assessment does not make the decision for anybody,” he said. “You still have professionals and human beings that are ultimately making and implementing the decisions.”

One of those professionals is Daniel Vokins, who has been a judge in Johnson County for more than 13 years. The score and the information provided by the risk assessment, he said, are a far cry from years past when defendants had to rely on defense attorneys to pass on information — that is, if they could even afford to hire one before their first court appearance.

“We didn’t have any information that zeroed in on what was going on in the defendant’s lives as far as work and home life,”  Vokins said. “This tool gives us a lot more information, not only when they are in court for the first time, but even when I review cases to decide whether to issue arrest warrants or not.”

The results of the risk assessment help Vokins choose whether to assign a cash amount for bail or to let a defendant go on a personal recognizance bond — a signed promise to come back for trial. The facts provided also inform whether he’ll impose other pretrial conditions on a defendant, such as drug counseling, alcohol monitoring or house arrest.

Vokins said pretrial detention and high bail amounts are more appropriate for people who are charged with crimes like murder, sexual assault or repeated DUIs. People charged with low-level crimes pose a much smaller public safety risk if released.

“Let people out that need to be out,” he said. “Help them with any services that might be available to turn their lives around.”

‘Garbage in …’

But some critics of data-based pretrial risk assessments say the formulas don’t do enough to mitigate the negative effects of pretrial detention.

“We shouldn’t be judging people and making decisions about their freedom based on what other people have done in the past,” said John Raphling, a senior criminal justice researcher at Human Rights Watch.

Those critics say risk assessments can perpetuate racist and class-based biases built into the data. Because people of color and poor people are already more likely to be stopped by the police and arrested, Raphling argued, the data will reflect that they are more likely to re-offend or skip court if they’re let out on bail.

“The arrest history that’s used to estimate whether you’re going to commit a future crime or not is already biased out the gate,” Raphling said. “The theory is: ‘garbage in, garbage out.’”

He argues in favor of more individualized attention for each person who is arrested.

“There should be real due process,” Raphilng said, “not some hokey machine that the estimates the likelihood of them coming back to court.”

Holsinger, on the other hand, argues that degree of individual attention is impossible in a crowded court system.

“It’s not a therapy session,” he said. “Justice systems are typically very busy systems.”

He said he controlled for demographics in his statistical analysis, isolating variables affecting defendants’ likelihood of skipping court or committing another crime, regardless of characteristics such as race or sex.

Data provided by Johnson County show that between March 2016 and May 2017, 15 percent of white defendants were assigned a “low” risk level, 28 percent were assigned a “moderate” risk level and 57 percent were assigned a “high” risk level. Those proportions were almost exactly the same for black defendants, the only other racial group for which data was provided.

Next steps

Racial bias, public safety and reducing jail and prison populations will be on the minds of the members of Kansas’ Pretrial Justice Task Force as they develop their recommendations over the next year and a half. Data-based pretrial risk assessment is likely to be on the list of recommended reforms.

“There’s a wide range of approaches,” said Arnold-Burger, chair of the task force. “And we need to find out what works best in our state.”

Meanwhile, Holsinger has plans for another revision of Johnson County’s pretrial assessment.

After a few years, the data showed that a history of mental and behavioral health issues wasn’t as good a predictor of negative outcomes as Holsinger thought it would be. He plans to remove the mental health flag from the list of risk factors associated with skipping court and committing another crime.

“That’s a part of the ongoing evolutionary nature,” he said, “of this entire endeavor.”

Nomin Ujiyediin is a reporter for the Kansas News Service, a collaboration of KCUR, Kansas Public Radio, KMUW and High Plains Public Radio covering health, education and politics. You can reach her on Twitter @NominUJ.

Kansas News Service stories and photos may be republished at no cost with proper attribution and a link to ksnewsservice.org.