Become a Patreon!
Excerpted From: Christopher Slobogin, Preventive Justice: How Algorithms, Parole Boards, and Limiting Retributivism Could End Mass Incarceration, 56 Wake Forest Law Review 97 (2021) (332 Footnotes) (Full Document)
When it comes to criminal justice, change is in the air. From calls to Defund the Police and Abolish Prisons to the Eight Can't Wait and Smart Sentencing initiatives, a cacophony of proposals for reorienting policing, pretrial detention decision-making, and sentencing are jumping out of academic journals and into mainstream political culture. This Article adds to the reformist hubbub by proposing dramatic changes to our system of punishment. It calls for a system of “preventive justice” that harks back to the days of sentences determined by parole boards, but with two important twists: sentence ranges would be consistent with retributive principles, and release would be required at the expiration of the low end of the range unless the offender is found to pose a high risk for committing violent crime, based on the results of a statistically derived risk assessment tool.
The prescriptions advanced here will undoubtedly strike many would-be reformers as the opposite of a reform agenda--a throwback to worn-out ideas and a dangerous endorsement of flawed, biased, and mechanistic technologies. The burden of this Article is to persuade otherwise.
Meeting that burden begins by emphasizing that a primary goal of preventive justice is to put a significant dent in our incarceration rates and massive prison populations. It is well-known that the imprisonment rate in the United States has skyrocketed since the late 1960s, from the neighborhood of one hundred people per one hundred thousand, to somewhere between five to seven hundred people per one hundred thousand, so that now prisons and jails house well over two million individuals. Although prison growth has moderated somewhat in the past several years, the pace of contraction has been slow. By one estimate, at the current rate of downturn, it will take until 2101 for the prison population to return to its 1980 level.
A growing number of policymakers and commentators think that postconviction incarceration rates are an acute problem. Those on the left are most concerned about the human cost, not only to suspects and offenders but to their families and their communities. In particular, these critics point to the huge proportion of people of color who are in prison--at a rate roughly six times that of whites to the disruption that imprisonment causes to the families and neighborhoods of those who are confined. More generally, critics emphasize the research that suggests that imprisonment is itself criminogenic. Those on the right may have these reactions as well, but are probably at least as concerned about the cost of prisons, jails, and correctional staff. The money spent on the carceral state has quadrupled since the 1990s, to upwards of $80 billion a year, and the indirect costs of this prison boom have been estimated at over $500 billion (about 6 percent of the U.S. gross domestic product). From a more philosophical perspective, both those who believe punishment should focus on just desert and those who are more utilitarian in orientation favor some degree of decarceration. Whatever the reason, a consensus has built among policymakers that reducing the number of people in prison is a critical objective, at least if it can be done without increasing the danger to the public.
The public seems to agree. In a 2006 poll, only 38 percent of respondents said that reducing the prison population was “very important” as a stand-alone goal, while 81 percent stated it was “very important” to ensure that “the punishment fits the crime.” But by 2016, surveys showed that over 80 percent favored reduction of prison populations as a primary goal of the criminal justice system. And when given more context, the public's attitudes toward incarceration appear to be even more attuned to its negative aspects. In one 2012 PEW poll, 78 percent of respondents stated that it would be acceptable to reduce prison time for low-risk, nonviolent offenders to close budget deficits, and over 80 percent believed that more money should be spent on alternatives to prison for such offenders. Well over a majority of respondents in the same poll endorsed the following statement: “It does not matter whether a nonviolent offender is in prison for 18 or 24 [or] 30 months .... What really matters is that the system does a better job of making sure that when an offender does get out, he is less likely to commit another crime.” In two polls of Wisconsin citizens (a “purple” state) in 2012-2013, a majority of respondents supported early release for both nonviolent and violent offenders--at a point halfway through the sentence no less--if the offender “can demonstrate that he is no longer a threat to society.”
Several solutions to the mass incarceration problem have been proposed and, in some states, have been implemented. Prominent initiatives include shortening sentences either at the front end or through early release, eliminating mandatory sentencing, creating more alternatives to jail and prison, and supporting funding initiatives prohibiting the use of state prisons for certain categories of offenders, which in California created an incentive for localities to develop cheaper options. More radical proposals--such as those subsumed under the Abolish Prison rubric--are unlikely to go anywhere, but in today's protest-energized world are more than just an academic pipedream and can at least provide baselines against which to measure more modest reforms.
As both the poll data and the experience of reform states indicate, however, substantial change in incarceration practices will not occur unless politicians, government officials, and the public have reasonable assurances that most of the individuals who are meant to benefit from these types of reforms will not commit new serious felonies. Even advocates for abolishing prisons always express the important caveat that confinement needs to be retained for the “dangerous few.” Those focused on civil liberties have an additional concern. They worry that, to the extent these initiatives depend upon the subjective judgments of judges and parole boards, the move toward discretion-based sentences and intermediate dispositions will not be evenly distributed, but rather will disfavor people of color or others who fit certain stereotypes.
The key thesis of this Article is that risk assessment instruments (“RAIs”)--statistically derived algorithms that estimate the risk of reoffending posed by groups of offenders--can play a significant role in assuaging both of these fears. RAIs can help differentiate high-risk and low-risk offenders while at the same time constraining the decisions that do so. In apparent recognition of these possibilities, about half the states use RAIs in some fashion at the dispositional stage or as a means of allocating correctional resources.
Despite their increasing prevalence, however, the full impact of risk algorithms has yet to be either realized or adequately assessed. That is because their use is, in every jurisdiction, entirely discretionary. In Virginia, for instance, a 2016 study found that RAI-based sentencing recommendations were overridden in 40 percent of the cases; judges imprisoned 42 percent of those recommended for alternatives to prison and permitted alternative dispositions for 23 percent of those recommended for imprisonment. Other researchers report similar dynamics.
From a decarceration perspective, that is unfortunate. Researchers with bipartisan credentials who audited the compositions of the prison populations in three states estimated that, if danger to the community were the only justification for continued confinement, roughly half the prisoners would be released. The Prisoner Assessment Tool Targeting Estimated Risk and Needs (“PATTERN”), the RAI developed in connection with the federal First Step Act, adopts a very narrow definition of low risk, but it nonetheless initially identified 48 percent of its sample population of prisoners as “low” or “minimum” risk and thus eligible for early release from their prison sentences. In Virginia, RAIs designed to recommend to judges who should be sentenced to prison alternatives identified 63 percent of drug offenders and 43 percent of larceny and fraud offenders as being low risk, with low risk defined as a 10-15 percent chance of recidivism within the next three years.
When offender risk and base rates for reoffending are not quantified in this way, the tendency, at least in the United States, is to opt for incarceration. In particular, judges and prosecutors who are subject to election (which describes almost all state court judges and prosecutors) and parole board members who owe their livelihoods to fickle politicians have good reason to avoid appearing “soft” on crime, given American cultural proclivities. As David Ball's parsing of the psychology literature shows, there is also “the human tendency to desire certainty and simplicity;” Ball suggests that this desire, which he attributes to judges as well as everyone else, “may help explain why [our] default seems to be to keep someone locked up, 'just in case'--and why this desire is resistant to information and argument.”
In contrast, the quantified results of well-validated RAIs can provide a concrete, rational basis for diversion or release. If, as recommended in this Article, adherence to those results is required in most circumstances, the human urge to incapacitate those in the law's grasp can be even more effectively resisted because decision-makers must obey the objective facts. Evidence of such a dynamic comes from Virginia, which found that judges using risk algorithms were willing to reduce sentences even for sex offenders “when they can point to the low risk-assessment score as a second opinion to support their decision.”
For the same reasons, racial and other types of bias in decision-making about postconviction release can be significantly reduced if the relevant cut-points have the force of law. Even if, as some claim, RAIs are more likely to misclassify black people than white people as high risk (a claim that this Article looks at closely), large numbers of black people will still be classified as low risk. If that categorization creates a presumption against incarceration, more people of color will be eligible for release. In contrast, a regime that is based on intuitive or clinical judgments about who is “dangerous” is too easily manipulated and prone to overly conservative outcomes influenced by conscious or unconscious prejudices. Indeed, the available research indicates that racial disparity becomes pronounced when, contrary to the recommendation of this Article, legal decision-makers depart from the risk algorithm.
RAIs can also help identify ways of ameliorating the risk of offenders, whether they are released or confined. The old mantra that “nothing works” in the battle against recidivism has been soundly debunked. As one 2008 review of the research summarized it: “The global question of whether rehabilitation treatment works to reduce recidivism has been answered in the affirmative by every meta-analyst who has conducted a systematic synthesis of a broad sample of the available experimental and quasi-experimental research.” Ten years later, another meta-analysis of institutional programs aimed at reducing the risk of adults confirmed that the recidivism of adults was roughly 35 percent lower than those who are not treated. In many types of cases, community-based programs are even more effective at curbing violent behavior. RAIs that identify criminogenic needs can link individuals to the appropriate programs.
In theory, then, RAIs could bring significant benefits: First, fewer people-- of all ethnicities--would be put in jail prior to trial and in prison after conviction. Second, for a substantial number of those who are imprisoned, overall sentences would be shorter. Third, treatment resources would be more efficiently allocated. Fourth, for these three reasons, the capital outlays for prisons and jails would be substantially less (although alternatives to prison, including good treatment programs, would cost more). Fifth, the public would continue to be protected, and perhaps would be even better protected, from the most dangerous individuals. Sixth, lower-risk individuals would be less subject to the well-documented criminogenic effects of incarceration and better positioned to build and maintain a life outside of jail or prison that does not involve criminal activity.
RAIs cannot fully realize these benefits, however, unless the currently popular determinate sentencing structure that exists in most states is dramatically altered. Today, determinate sentencing states give almost all sentencing power to prosecutors, who in essence fix the sentence range through charging practices, and judges, who decide where within the range the sentence will fall and occasionally select a sentence outside that range. Even in states that technically retain parole, the power of parole boards to affect sentences is very circumscribed. The result is that even an offender who poses a low risk of reoffending will often receive a lengthy prison sentence.
Preventive justice--again, a sentencing scheme that limits the prosecutor and judge to a determination of a sentence range, with the ultimate sentence determined through risk assessments--would restore power to parole boards. The indeterminate sentencing regimes of yesteryear, featuring broad sentence ranges and release decision-making by politically appointed parole board members, were much maligned, justifiably so in many respects. Critics voiced concerns about the incompetence of parole officials and the inaccuracy of predictions, among other objections. But if instead, parole boards were composed of experts in risk assessment, used RAIs in their assessments, and were required to release all but high risk offenders at the end of the minimum sentence demanded by desert, a much different type of indeterminate sentencing would exist--one that would dramatically reduce the amount of time offenders spend in prison, if they go to prison at all.
Even less refined parole schemes have produced such results. Kevin Reitz has made the argument that much of our prison growth since the 1990s has been in states with indeterminate systems controlled by parole boards, which increasingly refused to release prisoners who were eligible for parole. But he does not mention the fact that the same pressures that led to truth-in-sentencing, mandatory minima, and three-strikes laws also made “indeterminate” sentencing regimes much more determinate; legislative initiatives in these jurisdictions significantly increased the proportion of prisoners who are ineligible for parole, to anywhere from 45 to 93 percent. And while parole boards are certainly vulnerable to political pressures from governors and legislatures, that malleability works both ways; Reitz's work acknowledges that before the tough-on-crime movement in the 1980s and 1990s, parole boards were much more willing to release prisoners early, and that today the states that are leading the charge in the decarceration movement are those with indeterminate sentencing. Expert parole board decision-making constrained by RAIs should be able to do even better.
Part II of the Article provides more detail on this preventive justice regime and illustrates how it both resembles and differs from the indeterminate sentencing systems of yore. Parts III and IV grapple with the controversies surrounding RAIs; although some attention will be paid to complaints about the accuracy of these tools, the bulk of the discussion will be about algorithmic fairness, which has been the focus of most criticisms of RAIs. Part V calls on states to experiment with preventive justice approaches. The hypothesis of this Article, which needs to be given a fair test, is that a system of preventive justice offers the single most potent, and most realistic, mechanism for bringing about significant reform of the American criminal punishment system.
[. . .]
Risk and needs assessment instruments are crucial tools for pinpointing the hundreds of thousands of offenders who can, with relative safety, be diverted to community programs or be released with no restrictions. Without the quantitative clarity and authority of these instruments, governments will have neither the wherewithal nor the will to make serious inroads on our incarcerated populations. Given retributive urges, decriminalization will at most affect the lowest-level misdemeanors. For the same reason, significant reductions in sentences for more serious crimes are unlikely to be countenanced by the American public unless those reductions take place on an individualized basis and can be shown to be of lower-risk offenders. And without concrete proof that particular offenders are low risk, our elected and politically appointed decision-makers are, understandably, unlikely to opt against confinement.
In short, if the goal is to make significant inroads on the incarcerated population in the United States, risk assessment technology may be the only realistic method of doing so. RAIs would have an even greater impact if they are given presumptive effect. And they should have greater impact still if they are used within a system of preventive justice that relies on risk to calibrate the nature and length of sentences within a retributive framework.
Many jurisdictions are already using RAIs. But very few vet those instruments through a peer review process, give their results presumptive effect, or train judges, lawyers, and correctional officials in their use. Until a few pioneering jurisdictions give RAIs a fair shot, we cannot know whether the hypotheses that this Article advances about their benefits will be borne out. If the integration of well-constructed and presumptively applied RAIs into postconviction settings does not significantly reduce incarcerated populations, recidivism rates, and prison costs, it is probably not worth pursuing on a nationwide basis. But the bet here is that, without such experimentation, the American carceral system will be stuck where it is now for some time to come.
Become a Patreon!