How "Risk Assessment" Tools Are Condemning People to Indefinite Imprisonment

How “Risk Assessment” Tools Are Condemning People to Indefinite Imprisonment

Risk assessment tools used to evaluate prisoners are no more fair and free of bias than the societies that created them.Risk assessment tools used to evaluate prisoners are no more fair and free of bias than the societies that created them. (Image: Jared Rodriguez / Truthout)

Brad Lieberman has had more than his day in court. Starting in 1979 with his first arrest, he has had almost a dozen court dates, and counting. Lieberman was convicted in 1980 for seven rapes.

He was scheduled to leave prison in 1999. But today, 17 years after Illinois was required to release him, he is still locked up. If he had been convicted of murder, not sexual assault, he’d most likely be out. Why?

While eligible for release in 2000, then Illinois Attorney General Jim Ryan successfully petitioned to hold Lieberman under a 1998 Illinois law that permits a “sexually violent person” (SVP) to be indefinitely detained, or civilly committed, not through the Department of Corrections but through the Department of Human Services. Under this law, and similar ones now in 20 states and the federal system, an SVP is not a criminal but can be involuntarily confined, for treatment.

In 2007 Lieberman unsuccessfully petitioned to be released, providing testimony from a psychiatrist to challenge his diagnosis by the state’s psychologists of “Paraphilia not otherwise specified, sexually attracted to nonconsenting persons,” often shortened to “Paraphilia NOS, nonconsent.” (The dictionary defines “paraphilia” as a “condition characterized by abnormal sexual desires, typically involving extreme or dangerous activities”).

In 2016 psychological experts again dueled and on February 29 Lieberman’s most recent appeal for release was rejected. “The fact is that all the testimony before the court is that the condition he had doesn’t go away with time,” stated Judge Dennis Porter in a downstate Illinois courthouse.

Why did Judge Porter deny his release? Lieberman’s “mental abnormality.” The key evidence supporting this diagnosis? His convictions for rape, 36 years ago.

When 19 and 20 years old, Lieberman harmed a lot of people. If released, he could harm again. Is this possibility grounds for incarcerating him well beyond the end of his sentence?

While locking people up for potential crimes might sound like a middling Tom Cruise movie, or a new FX series, this future is now. Through predictive policing, criminogenics and risk assessment, our current wave of criminal justice reform argues that we can identify who is dangerous, who is likely to break the law. Risk assessment is increasingly used to determine who will be released on bail, who will be sentenced to prison, who will be granted parole and who will be kept on supervision once released.

And nowhere in the justice system is risk assessment more entrenched than in the foggy world of people with convictions for sex offenses. Just ask the more than 550 people, including Lieberman, held at Rushville Temporary Detention Center in southern Illinois, or the approximately 5,400 other people in similar facilities across the US.

While locking up people with convictions for sex offenses for years beyond their legal sentences may sound appealing to people seeking to combat sexual violence in our society, in truth, post-release punishments for people with convictions for sex offenses have done nothing verifiable to reduce rates of sexual violence. There is no research that suggests that sex offender registries and the multiplying regime of community notification laws, for example, have reduced child sexual violence, caught perpetrators, or protected children: Research supported by the US Department of Justice concludes that Megan’s Law, a key federal law that contributed to the establishment of sex offender registries, “has no effect on reducing the number of victims involved in sexual offenses.” States with civil commitment do not have lower rates of child sexual violence than states without civil commitment. Labeling the passports of people with convictions for sex offenses, or limiting their housing options and employment opportunities, does not reduce child sexual violence.

Yet, the fact remains: Most of the people in civil commitment facilities, like Lieberman, sexually assaulted children or women. If released, some of these folks might harm again. Others will not. But regardless of that uncertainty, once they have completed their sentences, is it acceptable for our society to use a checklist, a psychological evaluation, or a software program to legitimate continued confinement?

Delving into the world of risk assessment for those with convictions of sex offenses offers a window into the wider carceral reform movement’s increasing reliance on risk assessment — and it also stands as a warning.

A crucial question for this political moment is this: Will a turn to risk assessment shrink, or augment, the nation’s bloated prison and policing systems?

Science of Predictability

“Predictability is feasible,” wrote sociologist Ernest Burgess, one of the founders of predictability studies, or risk assessment, in 1928. Burgess, co-author of the 1939 Predicting Success or Failure in Marriage and a lifelong bachelor, created a scientific instrument, minus any “index factor” of love, to predict whether a (heterosexual) marriage would be “successful.”

Before his foray into marriage-success evaluation, Burgess created a “21 factor test” in Illinois to assess prisoners’ readiness for parole, asking, If released, is this person likely to reoffend? By 1933, during an era when “corrections” departments were at least nominally more focused on rehabilitation, Burgess’s instrument was in widespread usage in Illinois, and by 1939 predictability instruments shaped parole across the US, according to political science professor Bernard Harcourt in his book Against Prediction: Profiling, Policing, Punishing in an Actuarial Age.

The basic framework for most criminal justice predictability instruments is similar: Identify a group of incarcerated people as they are being released from prison. Gather data about their lives — generally the “static” or fixed factors, such as age, type of convictions, martial status, employment, education — and track who reoffends. Use the data points from the individuals’ lives that correlate with recidivism to form an instrument. Notably, recidivism is typically framed as highly individualized and without a context. Erased are the effects of formal and informal penalties that deny people with criminal records access to employment, housing and education and the tight requirements attached to parole.

By the 1980s and 1990s, the need for risk assessment diminished with laws that increased time behind bars and eliminated any discretionary power of judges. “Law and order” policymakers, often Republicans, challenged the belief that corrections could be rehabilitative and derided the supposed softness of parole, instead championing the slogan: Do the crime, do the time. Yet a range of liberal policymakers, including then Senator Edward Kennedy, also pushed for sentencing reforms — often referred to as “Truth in Sentencing” and partially responsible for our now ballooned prison population — to remove bias from the sentencing process, as African American Studies scholar Naomi Murakawa outlines in The First Civil Right: How Liberals Built Prison America.

Today, the pendulum of reform, and the role of risk assessment, is again in motion. Pre-trial release decisions will be informed by “data-driven, validated, pre-trial risk assessment tools,” according to a 2016 press release launching the Obama administration’s”Data-Driven Justice Initiative.” Proposed federal legislation with the capacity to reduce the federal prison population by 60,000 (the Sentencing Reform and Corrections Act) relies on risk assessment. Predictive or preemptive policing is simply “the same policing just faster and with more math,” says Jamie Garcia, a member of the coordinating team for the Stop LAPD Spying Coalition, and is still reliant on the misguided theory of deterrence. In LA, reports Garcia, Operation Los Angeles Strategic Extraction and Restoration (LASER)uses a risk assessment program to profile “chronic offenders” in an “identified hotspot.” State reform initiatives also often rely on risk assessment. For example, in 2015, Illinois acquired the Service Planning Instrument to “standardize” risk assessment across probation, parole and re-entry statewide.

Yet, how accurate are these instruments? Several main critiques have emerged. Many of these instruments, as some investigations have uncovered, aren’t that effective. They aren’t bias free: If an address “predicts” crime, couldn’t this be because specific (and almost always non-white) neighborhoods are under more surveillance by police than others? And while actuarial data may identify which life factors facilitate recidivism in a specific cluster or group, tools generated from this data appear less useful when applied to specific individuals.

ProPublica recently examined over 7,000 people in Florida’s Broward County who were assessed after arrest with one of the most widely used risk assessment programs in the country, Correctional Offender Management Profiling for Alternative Sanctions (COMPAS)created by the for-profit company, Northpointe. Out of every five people Northpointe predicted would reoffend, ProPublica’s analysis found that only one actually did so. And, notably: “The formula was particularly likely to falsely flag black defendants as future criminals, wrongly labeling them this way at almost twice the rate as white defendants.”

While Northpointe, citing propriety, refused to make public the analytics or algorithms, COMPAS includes the following questions:

“Was one of your parents ever sent to jail or prison?”

“How many of your friends/acquaintances are taking drugs illegally?”

“How often did you get in fights while in school?”

“A hungry person has a right to steal.” (Agree or Disagree)

Similarly, Operation LASER, according to Garcia, is shrouded in secrecy: “the algorithms that are used to create these hotspots are not transparent to the public and open for public scrutiny,” and, unsurprisingly “as police largely target communities of color and poor communities making the appearance of higher crime in those areas, historically biased and racist data is going into these algorithm and hotspot techniques.” Offered as “race neutral,” these technologies mask how race continues to shape all facets of the criminal justice system.

ProPublica is not the first to name problems with the ascendency of risk assessment. “The New Science of Sentencing,” a 2015 feature piece for The Marshall Project, critically explored the use of risk assessment tools in sentencing reform. In 2016, the American Civil Liberties Union raised questions about the for-profit company Intrado’s software, “Beware,” which creates profiles of individuals for police usage.

While the use of risk assessment instruments in sentencing and police profiling has garnered some attention from media outlets and advocacy groups, for those with convictions for sex offenses, where the stakes are arguably higher, scrutiny is almost nonexistent.

Civil Commitment

People with convictions for sex offenses can be detained, even after their sentence ends, based on their “mental abnormality,” according to the 1997 decision in Kansas v. Hendricks. Detention is not criminal but administrative, the court reasoned, and the goal is treatment, not punishment. This decision, and two subsequent Supreme Court cases, opened the door to civil commitment for anyone the state classifies as a sexually violent person/predator (SVP) or a sexually dangerous person (SDP).

The process for civil commitment across the states is fairly similar. For many people with a conviction for a sex offense the attorney general, or another state entity, triggers a review process at the end of their prison sentence. If, in the expert option of the evaluators hired and approved by the state (most require two evaluators), the person has a “mental abnormality,” a trial proceeds. (If not, the person is scheduled for release). A trial is held to determine whether the person meets the criteria of an SVP. If the process is contested from the start, a probable cause hearing can be held to determine whether there are grounds to hold the person pending trial, but these are almost always won by the state, according to Dr. Brian Abbott, a licensed psychologist with 38 years of experience working with people who have convictions for sex offenses.

Evaluators — generally psychologists licensed by the state — play a key role in the civil commitment process. Increasingly, these professionals rely on the mostly widely used instrument in the world to assess the likelihood that a person with a conviction for a sex offense will reoffend: a 10-question Yes/No checklist.

Called the Static-99R after the year it was created — 1999 (revised in 2012) — and for the “static” or fixed life-factors that predicted reoffending, Karl Hanson and David Thornton initially sought to review existing research to create a tool to house people with conviction for sex offenses more effectively in prison, and to decide who should be eligible for release.

Lower scores on the Static-99R (which has a maximum score of 12) supposedly indicate a lower likelihood of committing another sex offense. The tool focuses on the person’s convictions; the most points are allocated for prior offenses. Points are added for multiple convictions (particularly for sex offenses), and if the victims are unrelated to the perpetrator, are strangers and/or are male.

Does the Static-99R predict whether someone with a sexual offense might harm again? Fundamentally, according to Dr. Brian Abbott, “social science research doesn’t have this ability,” as “sexual recidivism risk science is too imprecise.”

Similarly, Daniel Coyne, a Chicago lawyer with clients in civil commitment facilities, states that actuarial data are “wonderful” for bigger-picture data — projections, research and longitudinal studies — but most were never designed for individual predictions. Actuarial data, he says, “might say I am due for a heart attack, but does this mean I will have one?” When applied to individuals, risk assessment instruments in the sex offender world only have a 58 percent accuracy rate, states Coyne. “Not much better than a coin toss.”

Given that SVP laws exist, what should be the standard for civil commitment? How should instruments like the Static-99R be used? Perhaps unsurprisingly, experts don’t agree.

“Alice In Actuarial-Land: Through the Looking Glass of Changing Static-99 Norms,” a 2010 research paper in the Journal of the American Academy of Psychiatry and the Law, suggests that the “lure of quantification” advanced the Static-99R without clear evidence to suggest that scores effectively predict recidivism. While acknowledging that professional judgment is not necessarily any more effective, the authors “sound” a “call to caution” in the “weighty decisions involving an individual’s liberty and the protection of public safety.”

Meanwhile, Dr. Abbott cautions against the move to elevate clinical judgment, arguing that this approach is “weaker” than actuarial assessments. In his response to this article, “Throwing the Baby Out With the Bath Water: Is It Time for Clinical Judgment to Supplement Actuarial Risk Assessment?” Abbot writes that unlike the murkiness of clinical judgment, the use of risk assessment instruments like the Static-99R brings “transparency, accountability, and consistency to the judicial risk-finding process in SVP/SDP proceedings that is not possible when using clinical judgment.” Or, at least an actuarial instrument can be held to some transparent scrutiny. Can the same be said about an expert’s professional judgment?

Even when risk assessments are used, professional judgment plays into decisions to confine or release someone. Take Lieberman: The determination to keep him confined was the result of both a “high score” on the Static-99R (largely thanks to his convictions of 36 years ago) and a “mental abnormality” that needs treatment, according to a professional. Lieberman’s diagnosis: Paraphilia not otherwise specified, nonconsent. A frequent diagnosis for SVPs, this diagnosis is also divisive.

“Paraphilia not otherwise specified” is in the “bible” of psychiatric disorders, the Diagnostic and Statistical Manual of Mental Disorders (DSM)-V. Yet, the “not otherwise specified” (or “NOS”) appendix, attached to other diagnosis in the DSM, is often criticized for vagueness. NOS designations are “residual wastebasket categories provided for clinical convenience” and should not be used for forensic purposes, wrote Dr. Frances Allen, a psychiatrist and former Chair of the Department of Psychiatry at Duke University School of Medicine who led the task force that produced the DSM-IV.

The picture gets sketchier. “ParaphiliaNOS, nonconsent” is not included in any version of the DSM. The American Psychiatric Association, which regulates the DSM, has repeatedly rejected the inclusion of ParaphiliaNOS, nonconsent (and some related precursors, including Coercive Paraphilia or “hypersexuality”). Why? It has stated that rape is a criminal act, not a mental disorder.

Wading into these boundaries, and negotiating potential turf wars between criminal and psychological realms, is tricky. The American Psychiatric Association declared in 1998 that the civil confinement of people with convictions for sex offenses was a “misuse of psychiatry,” yet the American Psychological Association has made no such proclamation.

Courts are reluctant to engage. Overall, in cases related to sex offenses, Coyne states that juries are often swayed by what he terms “the ick factor” and, when faced with scientific evidence that could help defendants, they either disregard this evidence or use it in service of further confinement. Coyne’s team’s current strategy is to raise how these instruments are not suited for individual predictions. The problem, however, is that accurate individualized assessment (and individualized treatment) “costs money.”

Meanwhile, we must wrestle with what is being fundamentally left out of risk assessment calculations. Despite their professed goal of securing the safety of the public, these risk assessment calculations completely neglect to include the needs and experiences of those who survive sexual violence — a population, studies show, that is rarely well served by the criminal justice system.

Also, sometimes, the line between perpetrator and victims that our system relies on is not so bright. Many of those who commit acts of sexual violence have themselves survived such violence in the past. Can a risk assessment demonstrate the capacity to heal from such violence? Should it?

In Limbo

David Jones is still waiting for his day in court. Now 30 years old, Jones has spent almost half of his life in institutions. Jones has been at Rushville for almost eight years without an SVP trial. His mother Edie Smith characterizes Rushville as in many ways “worse than prison” because there is no release date, and she feels her son, as a “sex offender” is often treated as less than fully human. (Smith and Jones are referred to by pseudonyms in this article due to Smith’s concerns for her son’s safety).

Smith isn’t convinced that Jones can be in the outside world yet, but she knows that Rushville is mostly harming, not helping, her son. She’s faced with the question of what to do with a loved one that needs more than the family can afford or offer.

Jones’s troubles started early. Grammar and middle school were pocked with school disciplinary actions, and Jones repeatedly sexually exposed himself to other young people. Putting the pieces together now, Smith sees patterns. When Jones was in his late teens she learned that other boys had extensively bullied him in school.

Smith doesn’t know why he was harassed. Maybe because he always looked a little different. He wore glasses. He didn’t have a lot of friends. Possibly a factor: He identifies as gay. She blames herself for the harassment: There was never enough money, she didn’t know enough to help him, his biological father was “not in the picture.”

Other young people also endure a litany of traumas, including schoolyard taunts, absent mothers and fathers, homelessness and sexual violence, but don’t respond by sexually exposing themselves. It is hard to know why Jones was different.

When Jones was almost 16, and the sexual exposure could no longer be dismissed as childish, a parole officer recommended Onarga Academy, a residential treatment center for juveniles with sex offenses. Smith’s first thought was money. How much would this cost? (While Onarga refused to release the cost, experts suggest it is easily over $100,000 per year). The options were jail and treatment, and Jones’s parole officer told her not to worry about the price tag.

Smith thought Jones was making progress but that he wasn’t ready to be out in the world, so she was “really pissed” when he was released after he turned 18. Onarga can house people until they are 21 years old, including those who are wards of the state.

After Onarga, Jones was unemployable. He was still “very immature emotionally,” says Smith, and he was viewed as toxic by potential employers, given his label of “sex offender.” Since Smith lived across from a school, he couldn’t live at home. He had nothing to do all day but “roam the streets.” Smith found a room for Jones in Chicago, but the landlord quickly found out he was a registered sex offender and asked him to leave. Smith struggled to find him another approved place to live within her nonexistent budget. She eventually found one, but at that point, Jones had exposed himself to another minor.

At 20, Jones was convicted for multiple misdemeanors related to sexual exposure, and for aggravated battery and criminal sexual abuse related to a sexual act with a 15-year-old male. (This second conviction, Smith states, was for what she reports was Jones’s consensual sexual activity with a minor.)

While a few people offered support, Smith and her family paid a heavy price when Jones was arrested as an adult. She reports that she was ostracized in her Illinois town. A neighbor told her that her son should be murdered. Committees formed to try to force her to move, even when her son was locked up (and therefore, obviously, not living at home). Neighbors and family members refused to talk to her.

Unlike four years earlier, when similar acts triggered a stay at a juvenile residential treatment program, Jones was sent to a medium-security prison, Pinckneyville. After serving his sentence, he wasn’t released: he was sent to Rushville.

Having now been confined through his 20s, Jones still waits. With a score of 11 on the Static-99R — he has multiple prior convictions, his victims were often unrelated, strangers and male (all “static” factors he will never be able to change) — he is a “high risk.” His risk assessment includes his juvenile convictions, even though the official coding manual cautions against the use of Static-99R with minors, or with adults who were convicted as minors.

Jones’s diagnosis? Smith is skeptical but reports that he was diagnosed in school with Attention Deficient Disorder, and at Onarga with Obsessive Compulsive Disorder, Bipolar Disorder and Mild Schizophrenia. Over the last eight years, reports Smith, the diagnoses have expanded to include a seeming laundry list of psychiatric disorders, including, “Paraphilia not otherwise specified, nonconsent.”

Smith wonders if another year or two at Onarga might have enabled him to develop enough to be able to live outside an institution. She also wonders why, at 20 years old, Jones’s life is assessed by the courts and the state as less valuable than it was when he was 16. Plus, why wasn’t even a fraction of the almost $100, 000 a year the state now spends to house him at Rushville made available to actually help Jones earlier on?

Risky Reform Futures

To many, the idea of predictability — an accurate forecast of who would do harm in the future, if released — is alluring in our widely unjust system. Risk assessments can appear more measured and bias-free than the judgment of police, judges, corrections officers or psychologists. The logic behind this reform wave: If police and prosecutors mistakenly target good people, an algorithm won’t. Risk assessment instruments are often seen as better than the status quo where few people convicted of violent offenses are released on parole or receive a reduced sentence.

But is the right question to be asking in this moment really how to balance professional judgment and a predictive instrument, or even how these algorithms or tools are inaccurate and racially profile? Instead, perhaps we should be asking why there is such a profound silence surrounding the elephant in the room: the structural problems in society that drive harm. Even if Burgess was right — that predictability is feasible — what meanings are drawn from these relationships? If reoffending can be predicted, why is the problem the person? Why not the wider environment that shapes our life pathways?

And why wouldn’t we conclude, as Harcourt writes in Against Prediction, “that there is a problem with prisons, punishment, or the lack of reentry programs”? And why wouldn’t predictability then motivate an examination of the root of these institutions, rather than a push to seek more reform?

Take the turn to predictive policing: Los Angeles already spends 54 percent of its general funds on its police department, says Garcia. With an inability to acknowledge and address racial bias within its force, continued histories of corruption, and a long list of lethal actions against unarmed residents, why isn’t the conclusion to shrink policing in LA, rather than produce new and expensive predictive tools that, according to a 2016 RAND study, don’t work?

Similarly, cities pay to expand facilities to monitor those with convictions for sex offenses, including civil commitment institutions, despite the fact that research has demonstrated that registries, and an array of other punitive responses, have neither protected children nor ended sexual violence. If our goal is to end, or reduce, sexual violence, why not focus and support proven preventative measures, including meaningful sex education and non-stigmatizing mental health care? Why not focus on ending patriarchy?

For many, this leap, for now, is too much. The lives and the rights of those with convictions for sex offenses, like Lieberman and Jones, are seen as necessary collateral damage in the wider march toward greater public safety, or perhaps exceptions to reform practices that on the whole will work for others. But as risk assessment in our political moment gains a measure of mainstream scrutiny, another story could emerge. Perhaps the focus on the supposedly incorrigible nature of people convicted of sex offenses (and the assumption that, given the gravity of their acts, many require indefinite detention) deftly masks our collective reality: Just like prisons, maybe big data can’t make us safe. In fact, it might make things worse.

Meanwhile, despite all these scores and diagnoses, Smith can’t predict her son’s future. With no money for a private lawyer, she does the best she can. She visits and calls and tries to stay informed about both what is happening at Rushville and how Jones is doing. She is pleased about the small things: Her daughter, who is now 18 years old, and therefore no longer a risk according to Rushville, can finally visit her brother. But for those in indefinite confinement and their families, the future is a question mark — and with big data in charge, the answer is often limitless incarceration.