July 17, 2017

Alexis Wilson Briggs
Director, Research and Development
Katal Center for Health, Equity, and Justice

Alec Karakatsanis
Founder & Executive Director
Civil Rights Corps

Dear Ms. Briggs and Mr. Karakatsanis,

Human Rights Watch writes in response to the bail reform recommendations spelled out in “8 Basic Principles for Money Bail Reform” (8 Principles). With the exception of Principle 4, we agree with each, and find them consistent with recommendations in our recent report, “’Not in it for Justice’ How California’s Pretrial Detention and Bail System Unfairly Punishes Poor People.” Other than Principle 4, the 8 Principles conform to international human rights standards. For example, Principle 7, which calls for decriminalization of “low level offenses,” echoes Human Rights Watch’s research and advocacy against criminal sanctions for crimes related to poverty, drugs, and mental illness. We have presented the human rights basis for opposing such forms of criminalization in past reports, including Nation Behind Bars, Every 25 Seconds: The Human Toll of Criminalizing Drug Use in the United States, and Callous and Cruel.

We recommend that Principle 4 be amended to urge jurisdictions to limit pretrial detention without the use of profile-based risk assessment tools. (We do not believe there should be any role for these tools in our pretrial justice system.) In fact, Principle 4 acknowledges problems with the use of these tools and primarily consists of admonishments for limiting and mitigating the harm that may come from their use. While we agree with the warnings implicit in Principle 4, we remain concerned that the recommendations offer no realistic constraints upon the harms described. For example, saying “such tools must be calibrated to take into account that people of color are disproportionately arrested, prosecuted and convicted,” is a laudable goal, but unlikely to actually occur. As Principle 4 states, “By their very nature, given current and past practice in policing, risk assessment instruments are prone to exacerbate racial disparities.”

Risk assessment tools make recommendations for critical decisions, including arrest, detention and supervision, based on a statistical prediction of the likelihood a person will do something bad in the future. This is an extremely troubling concept. It denies the individuality of the person about whom the decision is made and reduces him or her to a number. It ignores the specific context of that person’s life: a person with a “high risk” record may also have individual characteristics or needs that merit release without supervision or lower levels of supervision, but will be ignored by the prediction tool.

The prediction itself is based on data about how other people with similar characteristics have behaved in the past. In other words, the prediction is based on a profile. Because that data comes mostly from criminal history information, which we all know is highly skewed racially, in large part due to historical and on-going racial bias in policing, the profile itself is to a large extent racially determined. The critical release, detain or supervision decision should not be made based on a racially determined profile.

Principle 4 says the tools “must not re-entrench or mask discriminatory practices and disparities under the guise of ‘neutrality’ or ‘data…’” However, we know that pretrial detention greatly increases the chance of ultimate conviction, and that criminal conviction will increase the future risk assessed. So, a detention decision based on a racially biased risk assessment will feed back into future racialized decisions. Yet, these “validated,” “evidence-based” tools, by their very nature, do give a veneer of objectivity and science to the racialized prediction, making it seem more neutral and agreeable. We should reject anything that normalizes racial bias in our criminal justice (or any other) system.

Many proponents of risk assessment tools acknowledge this problem of racial bias. They argue, regardless, that in the existing system, judges make decisions based on implicit and explicit bias. This claim is true in many cases, but does not mean we should replace human bias with bias produced by algorithm.

Proponents argue that the risk assessment tools will facilitate release at much higher rates than the existing system. They point to the results in the first few months of New Jersey’s bail reform, and other jurisdictions that have seen significant reductions in pretrial detention rates. However, risk assessment is just one part of the pretrial reform measures in New Jersey or Kentucky, for example. Both have relied on vastly increased use of cite and release and other reforms as well to lower pretrial detention rates. Harris County, Ohio, which uses the Arnold Foundation risk assessment tool, has seen increased rates of pretrial detention and increased rates of early guilty pleas.

Risk assessment tools do not, in themselves, facilitate release. They are simply tools. The scoring system can be scaled, the algorithm adjusted, exceptions added or taken away, that will increase or decrease the number of people detained or placed under supervision. In New Jersey, these adjustments are already taking place, as political pressure mounts to keep more people in custody. While the initial adjustments may be relatively minor, greater adjustments may occur in the future. They could potentially even lead to higher rates of incarceration than existed before bail reform. The state court system, run by judges, has control over these adjustments, and judges have an institutional interest in pretrial detention to pressure rapid guilty pleas.

Even limiting the use of risk assessment tools is problematic. The idea that risk scores will be used only as one part of the decision-making process and that detention hearings will add individualized context places too much hope in our judiciary. Judges have an institutional interest in moving their cases rapidly. As reported in “Not in it for Justice,” judges tend to override release recommendations and follow detain recommendations, or they will simply default to the recommendation and ignore other information presented at hearings that humanizes a defendant. Full hearings slow courts down. Judges value speed, and following a risk assessment recommendation is faster than listening to a lawyer present individualized evidence and arguments for release. Rather than agree to using risk assessment tools, as advocates for justice, we should press for detailed, individualized and adversarial hearings with strict procedures and, as called for in Principle 1, strong legal standards that favor release. Of course, these hearings would only apply to the small number of people eligible for pretrial detention at all, as most should be issued a citation and released without arrest.

Even if the tools are limited to making decisions around release conditions, as is being proposed in California and in Principle 4, they have potential to increase onerous and invasive requirements like electronic monitoring, house arrest and probation reporting. While better than custody for most defendants, judges are likely to default to these types of conditions, or adjust the tools to increase their use. Racial bias from the tools in assigning release conditions is no more acceptable than racial bias in recommending detention.

We should adopt solutions to the inequities of money bail system that do not simply replace it with another biased system that may be used to over-supervise or over-incarcerate. As Principle 1 suggests, pretrial detention should be limited to an extremely narrow category of defendants. Legislation should be drafted to create bright line rules defining those narrow categories. In California, Human Rights Watch has recommended limiting the possibility of pretrial detention to those accused of “serious” or “violent” felonies, as defined in the Penal Code. The detention or bail hearing for the small eligible group should have strict standards requiring specific proof of an individualized danger or flight risk, before the court can consider setting any conditions of release or bail. Instead of using profile-based risk assessment tools, courts should assess risk through lengthy, detailed hearings in which the defendant has the resources to present mitigating evidence about their individual circumstances. Detention or release should not be determined by a statistical probability.

Our pretrial detention systems need to be changed. The “8 Principles” provide an excellent framework, except that their proposed limitations on risk assessment tools do not go far enough. Given the many harms that are likely to flow from their use, harms that are foreshadowed in the language of Principle 4 itself, we should oppose use of these tools entirely.


John Raphling
Senior Researcher