Skip Navigation

What Are Risk Assessments — and How Do They Advance Criminal Justice Reform?

Data can help us build a fairer system, but only if it’s used carefully

  • Vienna Thompkins
August 23, 2018

Too many people spend time in jail before they’re ever found guilty, merely because they’re suspec­ted of commit­ting a crime. 

Some states have tried to correct this prob­lem by using “risk assess­ments.” When a judge is consid­er­ing send­ing someone to jail ahead of trial, rather than setting bail — essen­tially tying someone’s free­dom to their abil­ity to pay for it — the judge will eval­u­ate whether the defend­ant is a public safety threat. If they are a threat, they’ll be sent to jail. If not, they get to go home. 

But how do you decide whether someone is a safety threat? That’s a harder prob­lem to solve. And that’s where risk assess­ments come in. 

What are risk assess­ments?

In short, risk assess­ments are a process or a tool the court uses to decide whether someone is too danger­ous to release, or a “flight risk” — that is, not likely to show up for trial. 

There are two types of assess­ments: clin­ical and algorithmic. In a clin­ical assess­ment, a psycho­lo­gist or clini­cian eval­u­ates someone indi­vidu­ally. These assess­ments can be time-consum­ing, requir­ing informal eval­u­ations, inter­views, and invest­ig­a­tions of vari­ous factors of a defend­ant’s life. 

An altern­at­ive method is an “algorithmic risk assess­ment.” Used to improve or completely replace clin­ical assess­ments, algorithmic assess­ments use data, stat­ist­ics, and social science to categor­ize a person’s risk level based on their responses to ques­tions or based on data about them. This typic­ally results in a numer­ical risk score and a desig­na­tion of low, medium, or high risk.

When are they used?

Risk assess­ments can help inform judges as they make decisions about pretrial deten­tion, bail, sentence length, and more. Some states have imple­men­ted these assess­ments at one or more of these stages. At the pretrial deten­tion and bail stages, an under­stand­ing of whether the defend­ant poses a risk for flight or viol­ence heav­ily informs the judge’s decision-making process. At the senten­cing stage, risk of viol­ence and recidiv­ism help a judge decide how long someone should be sent to prison. And another, similar type of assess­ment — a “needs” assess­ment — can help prison offi­cials decide what types of services, programs, and hous­ing a person will need during their time in prison. 

Are they useful? Do they have unin­ten­ded consequences? 

Risk assess­ments might help reduce the use of bail — which results in people going to jail because they are poor, not because they’re a public safety risk. Theor­et­ic­ally risk assess­ments might help end that prac­tice, making jail a place for people judged too danger­ous to be released rather than people too poor to pay to get out. 

But there are some reas­ons to exer­cise some caution. As noted by former Attor­ney General Eric Holder, algorithmic risk assess­ments are built upon data, and that data itself can reflect the systemic bias of the crim­inal justice system. For example, a prior arrest might make someone look “riskier” to an algorithm. But we know that people of color are more likely to be arres­ted than whites for the same crime. Despite rely­ing on neut­ral data, risk assess­ments might “bake in” a history of bias. 

Recent research suggests that these concerns are valid. More research can help determ­ine how to create effect­ive, unbiased risk assess­ments.

(Image: erhui1979/Getty)