Researchers found that data from the Client Legal Utility Engine (CLUE) system can be used to advance research, and fuel risk assessment-based audits of human decisions
The results of a $313K University of Maryland study funded by the Maryland Governor’s Office of Crime Prevention, Youth, & Victim Services are in: Data from the Maryland Volunteer Lawyers Service’s Client Legal Utility Engine (CLUE) tool can help put more of the “just” in criminal justice decisions.
Initially developed to “scrape” data from the state’s court case database to make it easier for Marylanders to find and expunge open cases, the CLUE system has since been used by criminal justice researchers to answer difficult-to-answer questions about trends in decision-making across the state’s 24 jurisdictions, and the impact of state reform efforts.
“Maryland is a siloed state in terms of its criminal justice data, and it's very hard to get insight into the decisions, policies and procedures that are happening across the state’s 24 jurisdictions. That makes the completion of this project even more of an accomplishment for this research team,” said Bianca Bersani, Director of UMD’s Maryland Crime Research and Innovation Center (MCRIC) and an associate professor in the Department of Criminology and Criminal Justice (CCJS).
The MCRIC team spent the last year turning CLUE system information about all Maryland counties’ 2012-2020 court cases into quantitative data points. Then, in order to validate the accuracy and quality of the CLUE data, they compared the pretrial detention outcomes of those court cases—whether someone was offered bail, released, and/or jailed—against the pretrial detention outcomes the Maryland Judiciary published to show whether a 2017 bail reform effort had its intended effect of keeping low-risk offenders out of jail by lowering, if not eliminating, overly expensive bail costs.
“We took those report numbers and then compared them with the numbers we came up with from all the jurisdictions in Maryland to see if we were getting numbers that were similar. And indeed, we were,” said Zubin Jelveh, the project’s lead faculty researcher, and an assistant professor in CCJS and the College of Information Studies (iSchool).
Once they validated the CLUE data, the researchers then asked themselves how the data could be used to improve what Jelveh believes was an unintended consequence of the 2017 bail reform effort—there being almost no noticeable reduction in the number people getting detained before trial, because they weren’t being offered bail at all.
“What we wanted to see was could we use CLUE data to assess the risk level of the people who were getting shifted from bail, to not getting released at all,” said Jelveh.
To do so, Jelveh and Emily Glazener, a CCJS faculty assistant and a fellow faculty researcher on the project, compared Maryland counties’ release/detention decisions to the forecasts of Arnold Ventures’ Public Safety Assessment (PSA), a pretrial risk assessment that Jelveh says some states are using. The researchers received mixed results. The researchers found that a large share of defendants who were detained after the reform were predicted to be at elevated risk of being rearrested for a violent crime by the PSA, suggesting that decision-makers were accurately taking risk into account when making detention decisions. On the other hand, the researchers also found little change in release rates for low-risk defendants, contrary to the intention of the bail reform.
“Even though a number of jurisdictions are using these tools to aid in pretrial decisions, the current evidence suggests that their impact has been modest at best,” said Jelveh. “However, you could use the output of a risk forecasting system like this to audit decision-making, rather than inform it. I think that's one of the big promises of data like this. The human brain is a black box, but we can use this data to see if the decisions they're making align with what policies ask them to do.”
Cracking into the human brain’s “black box” is something the researchers hope to dig deeper into. They also hope to use CLUE data to see how other risk assessments, like those that determine the likelihood of someone failing to appear for their hearing, could further inform and aid Maryland’s criminal justice reform efforts.
“Future research focusing on risk assessments in the state of Maryland is an important next step in this research,” said Glazener. “Multiple Maryland counties over the past decades have implemented established pretrial risk assessment tools from outside sources or generated their own tools, and there’s little understanding of how the introduction of these tools have impacted pretrial decision-making across the state.”
The impact of this work extends beyond Maryland, Bersani adds.
“Not only do the findings of this study speak to patterns statewide, they also enter into a timely discussion about pretrial processes aimed at helping to make sure that fewer people unnecessarily experience the negative outcomes of being detained, including criminal legal system outcomes, and the negative outcomes related to their employment status and a strained family life.”
The main photo is from iStock