At the start of every criminal case in New York City, a judge must make a critical decision: whether a defendant should be released before their trial, have bail set, or be detained. Historically, the judge decided based on a personal prediction of the defendant’s risk for not showing up for court. Yet despite this decision’s enormous impact, a large body of behavioral science research tells us that all people, including judges, have incredible difficulty accurately assessing probabilities of future events. In the courtroom, this thinking trap results in many people who were at low-risk for missing court ending up detained, while many high-risk defendants are released. Improving the accuracy of this crucial risk prediction could prevent thousands of jail admissions each year without threatening New York City’s reputation as the safest big city in the country.
Updating the New York City Criminal Justice Agency Release Assessment
New York City Criminal Justice Agency Release Assessment
The Crime Lab worked with the Mayor’s Office of Criminal Justice in New York City, the Criminal Justice Agency, as well as other city, state, and nonprofit partners to build, test, and implement a risk assessment instrument called the Release Assessment, which replaced NYC’s prior risk assessment that had been last updated in 2003. The Release Assessment is a data-driven decision aid that helps judges better identify how likely a given person is to return for all required court dates.
Before implementing the updated tool in courtrooms citywide, we tested the tool by seeing how it would have assessed past cases. We leveraged the latest data science to analyze information about more than 1.8 million cases in New York City between 2009 and 2015. By looking at patterns in past cases (which defendants successfully returned to court and what characteristics those defendants tended to have), the research revealed eight easily interpretable factors which, in combination, are most predictive of future court appearance. These eight factors (which do not include race, age, or gender) represent the algorithm that underpins the CJA Release Assessment. The Release Assessment went live in late 2019 and is currently being used to support judges in deciding on the conditions of a defendant’s pretrial release.
The Release Assessment increased both the accuracy and equity over the prior pretrial tool NYC was using. Under that prior tool, only 35% of defendants were recommended for release without conditions. The updated CJA Release Assessment recommends 89% of defendants for release. Moreover, whereas the prior tool had a substantial difference in the rate at which White and Black defendants were recommended for release without conditions, the new Release Assessment effectively reduced that gap to zero.
And because of the increased accuracy of this data-driven tool, the CJA Release Assessment can recommend far more defendants for release without any projected increase in missed court appearances. In sum, the CJA Release Assessment has the potential to help thousands of New Yorkers – in particular, people of color – avoid unnecessary pretrial incarceration at Rikers Island.