Among 34 pages of revisions to Missouri’s bond rules that took effect July 1, the phrase that has caused a stir is only 14 words long.

It appears at the end of a list of factors that judges “shall take into account” when deciding whether to detain or release a criminal defendant pretrial. The final factor judges must consider: “a validated evidentiary-based risk-assessment tool approved by the Supreme Court of Missouri.”

Such a tool is, in essence, an algorithm (or digital process) that receives a defendant’s information and then forecasts his or her compliance with conditions of release by referring to a database of the outcomes of similar defendants in the past.

Currently, St. Louis County is developing a tool of this kind that is modeled on the Public Safety Assessment, or PSA, developed by the Laura and John Arnold Foundation, according to circuit court spokeswoman Christine Bertelson. That philanthropic organization claims that the PSA, which already is in use of dozens of locations nationwide, draws on data from 750,000 cases in nearly 300 jurisdictions. It isolates for judges nine factors that can predict a defendant’s pretrial risk of failure to appear, new criminal activity and new violent criminal activity.

Bertelson said the PSA-style tool will be customized to its needs and should be finalized this summer, with training to begin in October.

But the mere prospect of these tools being deployed in Missouri has drawn criticism from outside the state.

“We expect to request all judges and pretrial services programs to cease and desist in their use of such tools,” said Jeffrey Clayton, executive director of the American Bail Coalition, a national trade association for insurance companies that underwrite bail bonds. Clayton wrote in a statement that these digital tools should not infringe on the right to cash bail, which “creates a check on the government’s power yet allows judges to achieve a delicate balance” in considering the rights of the various parties involved.

Another criticism came on July 17, when more than two dozen researchers from MIT, Harvard, Princeton, NYU and the University of California, Berkeley released a signed, four-page statement saying that pretrial risk-assessment tools are based on distorted data, false assumptions and “technical flaws” that raise “grave concerns.” Three of the researchers sent their statement directly to the Missouri Supreme Court.

“Pretrial risk assessments do not increase the likelihood of better pretrial outcomes, much less guarantee them,” the researchers wrote in an introductory letter to the high court. “The technical problems cannot readily be resolved — indeed, they are as much problems of theory as of implementation.”

These concerns may be premature: Despite the mandate that judges take it into account, no pretrial assessment tool has yet been validated or approved by the high court, according to Kansas City defense attorney J.R. Hobbs, one of three co-chairs of the Missouri Supreme Court’s Task Force on Criminal Justice. That body met during a period of two years and recommended a broad set of changes to the state’s bond rules, many of which were adopted by the high court in December and put into effect on July 1, with modifications to follow in January.

Hobbs added that the tool is “only one factor a judge will look at.” The other factors listed in the rules include the defendant’s family ties, employment, financial resources and mental condition; the nature of the charge; the weight of the evidence against the defendant; the defendant’s criminal record and record of appearance at other court proceedings; and whether he or she was on probation or release at the time the offense occurred.

Beth Huebner, a criminal justice professor at the University of Missouri-St. Louis who has been involved in efforts to reduce St. Louis County’s jail population, said the PSA-style tool could be an improvement, but its effectiveness depends on the quality of the data inputs and the training that judges receive.

“If you adopt the tool without any training or investment, things will not change,” Huebner said.

According to the Arnold Foundation, the PSA bases its forecasts on nine factors such as a defendant’s age, offense, prior convictions, prior sentence and prior failures to appear. “The PSA does not rely on factors such as race, ethnicity or geography,” the website states.

Yet the university researchers who criticized PSA-type tools claim that racial bias is baked into the algorithms because they rely on historical arrest-record data, which is not “neutral.”

The researchers argue that arrest records are, on the one hand, under-inclusive because most crimes don’t result in arrest, and on the other hand, they’re over-inclusive because “decades of research have shown that, for the same conduct, African American and Latinx people are more likely to be arrested, prosecuted, convicted and sentenced to harsher punishments than their white counterparts.”

Johnson County Prosecuting Attorney Robert Russell, who sat on the task force subcommittee that weighed introducing the tool into the new bond rules, said he and his colleagues had heard such criticisms during their deliberations.

“I think [the researcher critics] suffered from the perception that we were somehow endorsing the risk-assessment tool at the expense of all other considerations, and that’s simply not true,” Russell said. Judges aren’t bound to its suggestions, he emphasized, and he doesn’t believe judges will start defaulting to them, either.

Russell said that after hearing from a variety of entities, including academics and representatives of other jurisdictions that had implemented the PSA, he and his colleagues recommended that the high court consider approving that tool — but approval won’t happen overnight. The pilot project must train court officers and employees, set up the data infrastructure, then measure outcomes to verify, for example, whether defendants who are released on expectations of compliance do, in fact, comply. Once it’s validated, the Supreme Court must also sign off.

Predicted Russell: “I think we’re probably still several years off from that actually occurring.”

By Nicholas Phillips | Missouri Lawyers Weekly