In the Matter of Jorge L. Linares, Appellant,v.Andrea W. Evans,, Respondent.BriefN.Y.September 8, 2015APL-2014-00076 Albany County Index No. 5033/12 Court of Appeals of the State of New York IN THE MATTER OF JORGE LINARES, Petitioner-Appellant, – against – ANDREA W. EVANS, Chair of the New York State Board of Parole, Respondent-Respondent. BRIEF OF CRIMINOLOGY EXPERTS AS AMICI CURIAE BORIS BERSHTEYN RAMYA RAVISHANKAR SKADDEN, ARPS, SLATE, MEAGHER & FLOM LLP Four Times Square New York, New York 10036 Tel.: (212) 735-3000 Fax: (212) 735-2000 boris.bershteyn@skadden.com Counsel for Amici Curiae Brennan, Cullen, Holsinger, Koetzle, Kohler-Hausmann, Listwan, Travis, and Wormith Date Completed: July 22, 2015 ii TABLE OF CONTENTS Page IDENTITIES AND INTERESTS OF AMICI CURIAE .................................................................1 PRELIMINARY STATEMENT .....................................................................................................2 ARGUMENT ...................................................................................................................................5 I. TODAY’S SUBJECTIVE ASSESSMENT METHOD IS INEFFECTIVE. .......................5 A. The Board’s current subjective assessment method fails to adequately consider and reduce recidivism. ...............................................................................5 B. The subjective assessment method is susceptible to errors and opaque. .................8 II. NON-DISCRETIONARY USE OF COMPAS WOULD SUBSTANTIALLY IMPROVE NEW YORK STATE’S PAROLE DETERMINATIONS. ............................12 A. How COMPAS works ...........................................................................................12 B. Other states have succeeded with risk and needs assessment tools. ......................15 C. COMPAS outperforms the subjective assessment method. ...................................17 i. COMPAS better predicts likelihood of recidivism than subjective assessment. .................................................................................................17 ii. COMPAS increases consistency in parole decision making. ....................19 iii. COMPAS helps to eliminate implicit bias. ................................................19 iv. Decisions made using COMPAS rest on a more robust evidence base. ...........................................................................................................20 v. COMPAS promotes transparency. .............................................................20 CONCLUSION ..............................................................................................................................21 iii TABLE OF AUTHORITIES Cases Canales v. Hammock, 105 Misc. 2d 71 (Sup. Ct. Richmond Co. 1980) .................................................. 9 Coaxum v. New York State Board of Parole, 14 Misc. 3d 661 (Sup. Ct. Bronx County 2006) ................................................... 7 Franklin v. Shields, 569 F.2d 784 (4th Cir. 1977) .............................................................................. 10 Greene v. Smith, 52 A.D.2d 292 (4th Dept. 1976) ......................................................................... 21 Hawkins v. New York State Department of Corrections & Community Supervision, No. 0011-15 (Sup. Ct. Sullivan County, May 11, 2015) ................................ 9, 10 King v. New York State Division of Parole, 190 A.D.2d 423 (1st Dep't 1993), aff’d, 83 N.Y.2d 788 (1994) .......................... 7 Rabenbauer v. New York State Department of Corrections & Community Supervision, 46 Misc. 3d 603 (Sup. Ct. Sullivan County 2014) .......................... 7 Rios v. New York State Division of Parole, 15 Misc. 3d 1107(A), 2007 N.Y. Slip Op. 50529(U) (Sup. Ct. Kings County 2007) .............................................................................................. 7 Shepard v. Taylor, 556 F.2d 648 (2d Cir. 1977) ................................................................................. 7 Statutes and Rules Colo. Rev. Stat. Ann. § 17-22.5-404 (West Supp. 2014) .................................. 16, 17 N.Y. Correct. Law § 112(4) (McKinney 2014) ......................................................... 4 N.Y. Correct. Law § 805 (McKinney 2014) .............................................................. 4 N.Y. Exec. Law § 259-c(4) (McKinney Supp. 2015) ................................................ 2 N.Y. Exec. Law § 259-i(2)(c)(A) (McKinney Supp. 2015) ...................................... 3 iv Other Authorities Amy Robinson-Oost, Note, Evaluation as the Proper Function of the Parole Board: An Analysis of New York State’s Proposed Safe Parole Act, 16 CUNY L. Rev. 129 (2012) ......................................................... 15 Bryn A. Herrschaft, Evaluating the Reliability and Validity of the Correctional Offender Management Profiling for Alternative Sanctions (COMPAS) Tool: Implications for Community Corrections Policy (Jan. 2014) (unpublished Ph.D. dissertation, Rutgers University), https://rucore.libraries.edu/rutgers- lib/46260/PDF/1 .................................................................................................. 17 Christine S. Scott-Hayward, The Failure of Parole: Rethinking the Role of the State in Reentry, 41 N.M. L. Rev. 421 (2011) ................................. 16 Christopher Lowenkamp & Edward Latessa, Understanding the Risk Principle: How and Why Correctional Interventions Can Harm Low-Risk Offenders, in Topics in Community Corrections (2004) .................... 10 Christopher Uggen & Suzy McElrath, Parental Incarceration: What We Know and Where We Need to Go, 104 J. Crim. L. & Criminology 597 (2014) ..................................................................................... 11 D.A. Andrews & James Bonta, Rehabilitating Criminal Justice Policy and Practice, 16 Psychol. Pub. Pol’y L. 39 (2010) ............................................ 14 D.A. Andrews, James Bonta & J. Stephen Wormith, The Recent Past and Near Future of Risk and/or Need Assessment, 52 Crime & Delinq. 7 (2006) ........................................................................................ 8, 12, 19 Edward R. Hammock & James F. Seelandt, New York’s Sentencing and Parole Law: An Unanticipated and Unacceptable Distortion of the Parole Boards’ Discretion, 13 St. John’s J. Legal Comment. 527 (1999) ......................................................................................................... 6, 9 James Bonta, Offender Risk Assessment: Guidelines for Selection and Use, 29 Crim. Just. & Behav. 355 (2002) .......................................................... 14 James Bonta, Risk-Needs Assessment and Treatment, in Choosing Correctional Options That Work: Defining The Demand and Evaluating the Supply (A.T. Harland ed., 1996) .................................................. 9 v Joseph T. McCann, Risk Assessment and the Prediction of Violent Behavior, Fed. Law., Oct. 1997 ............................................................................ 6 Lin Song & Roxanne Lieb, Wash. State Inst. Pub. Pol’y, Recidivism: The Effect of Incarceration and Length of Time Served (1993) ......................... 11 Mary E. Pattillo et al., Russell Sage Foundation, Imprisoning America: the Social Effects of Mass Incarceration (2004) ............................... 11 Megan Comfort, Doing Time Together: Love and Family in the Shadow of the Prison (2008) .............................................................................. 11 National Reentry Resource Center, Reducing Recidivism: States Deliver Results (2014) ........................................................................................ 16 New York State Commission on Sentencing Reform, The Future of Sentencing in New York State: Recommendations for Reform (2009), http://www.criminaljustice.ny.gov/pio/csr_report2- 2009.pdf .............................................................................................................. 18 Northpointe, Practitioners Guide to COMPAS (2012) ...................................... 12, 15 Sharon Lansing, Division of Criminal Justice Services, New York State COMPAS-Probation Risk and Need Assessment Study: Examining the Recidivism Scale’s Effectiveness and Predictive Accuracy (2012), http://www.criminaljustice.gov/crimnet/ojsa/opca/compas_probati on_ report_2012.pdf ............................................................................................. 17, 18 Stephen D. Gottfredson & Laura J. Moriarty, Statistical Risk Assessment: Old Problems and New Applications, 52 Crime & Delinq. 178 (2006) .......................................................................................... 5, 18 Susan Turner & Julie Gerlinger, Risk Assessment and Realignment, 53 Santa Clara L. Rev. 1039 (2013) .................................................................... 6, 14 The Growth of Incarceration in the United States: Exploring Causes and Consequences (J. Travis, B. Western & S. Redburn eds., 2014) ................ 11 Tim Brennan, William Dieterich & Beate Ehret, Evaluating the Predictive Validity of the Compas Risk and Needs Assessment System, 36 Crim. Just. & Behav. 21 (2009) ................................................ 6, 9, 17 vi William M. Grove et al., Clinical Versus Mechanical Prediction: A Meta-Analysis, 12 Psychol. Assessment 19 (2000) .......................................... 8, 9 William M. Grove & P.E. Meehl, Comparative Efficiency of Informal (Subjective, Impressionistic) and Formal (Mechanical, Algorithmic) Prediction Procedures: the Clinical-Statistical Controversy 2 Psychol. Pub. Pol’y L. 293 (1996) .............................................. 18 IDENTITIES AND INTERESTS OF AMICI CURIAE Amici are criminology experts whose teaching and scholarship concerns criminal justice, behavioral sciences, and sociology. Pursuant to Rule 500.23(a)(4)(iii), they submit this brief because they expect it to be of assistance to the Court. Amici have a collective professional interest in the sound and consistent application of non-discretionary, evidence-based practices in New York Board of Parole decisions. This case presents an opportunity for the Court to render a decision that will reinforce the New York Legislature’s aim to improve parole decision making by incorporating non-discretionary risk/needs assessment tools, thereby affecting the lives of many in varied communities across the State. The experts joining this brief are:1 Tim Brennan, Chief Scientist, Northpointe Institute; Visiting Scholar, Georgia State University; Developer of COMPAS Francis T. Cullen, Distinguished Research Professor Emeritus of Criminal Justice and Senior Research Associate, The Center for Criminal Justice Research, University of Cincinnati Alexander Holsinger, Professor, Director of the Graduate Program in Criminal Justice, College of Arts and Sciences, University of Missouri-Kansas City Deborah Koetzle, Associate Professor, Executive Officer of the Doctoral Program in Criminal Justice, John Jay College of Criminal Justice 1 Institutions are provided here for identification purposes only. 2 Issa Kohler-Hausmann, Associate Professor of Law and Sociology, Yale University Shelley Listwan, Associate Professor, College of Liberal Arts and Sciences, University of North Carolina Charlotte Jeremy Travis, President, John Jay College of Criminal Justice Stephen Wormith, Professor and Director, Centre for Forensic Behavioural Science and Justice Studies, College of Arts and Science, University of Saskatchewan PRELIMINARY STATEMENT This case is about improving parole decision making. In 2009, the New York State Legislature responded to an extensive report by the New York State Commission on Sentencing Reform by inviting the optional incorporation of evidence-based practices to structure release decisions. The New York Board of Parole, however, did not adopt such practices as its decisional framework. In 2011, the Legislature acted more firmly by mandating the incorporation of evidence- based practices. Aiming to reduce recidivism and increase transparency, it directed the Board to create “written procedures for its use in making parole decisions . . . .” N.Y. Exec. Law § 259-c (McKinney 2011). The Legislature directed that these procedures “shall incorporate risk and needs principles,” as the method by which the Board would measure whether inmates presumptively meet the release standard. Id. (emphasis added). Board members were required to (1) use a 3 scientific method to assess an inmate’s recidivism risk and rehabilitation needs and (2) make their final decision based on a set of eight discretionary factors, which include an inmate’s institutional record and vocational education as well as employment and support services available to the inmate. N.Y. Exec. Law § 259- i(2)(c)(A). In response, the Board introduced COMPAS Reentry Risk and Needs Assessment Instrument (“COMPAS”), a validated risk and needs instrument that helps Board members objectively determine an inmate’s recidivism risk and craft a treatment plan based on the inmate’s post-release needs. The Board has not, however, established written procedures requiring its members to use COMPAS in the manner it was designed by its creators and intended to be used by the Legislature: to scientifically evaluate recidivism risk and rehabilitation needs. Instead, the Board has appended COMPAS to the eight discretionary factors from which the Board then makes an unstructured assessment. This discretionary use of COMPAS subordinates risk and needs principles to the very subjective risk assessments that precipitated legislative change. Without intervention from this Court, Board members will continue to make their decisions in a manner that lacks rigor and transparency and fails to reliably evaluate recidivism risk. Amici submit that COMPAS must be used in a non-discretionary manner in order for the New York State Board of Parole to meet its obligation to determine 4 whether “there is a reasonable probability that,” if an inmate is released, “he will live and remain at liberty without violating the law. . . .” N.Y. Correct. Law § 805 (McKinney).2 This probabilistic determination, based on many dynamic factors, calls not for unstructured discretion but for an “empirically validated” method required by the legislature. Id. at § 112(4). First, COMPAS has been found to accurately predict a parolee’s risk of recidivism in New York State. Indeed, studies demonstrate that COMPAS predicts recidivism better than the Board’s current subjective assessments. Second, COMPAS evaluates risk of recidivism and parolees’ needs after release based on objective criteria. Using objective criteria increases consistency across release decisions and reduces the biases inherent in subjective assessment. Finally, COMPAS provides a holistic, scientific method for evaluating the factors that the Board must consider to comply with Executive Law § 259-i. Thus, non-discretionary use of COMPAS is a prerequisite to accomplishing the Legislature’s objectives. 2 Amici endorse the use of validated risk/needs assessments in decision-making and do not promote one instrument over the other. Discussion of the COMPAS is included because the New York State Division of Criminal Justice Services adopted it as its risk/needs assessment in 2011. 5 ARGUMENT I. TODAY’S SUBJECTIVE ASSESSMENT METHOD IS INEFFECTIVE. Without a scientific tool like COMPAS, Board members deciding whether to release an inmate rely on their intuitive assessment of limited evidence. The typical method involves Board members, most of whom have little to no professional training in criminology or psychology, asking inmates non- standardized questions or using checklists to gauge and predict behavior. See Stephen D. Gottfredson & Laura J. Moriarty, Statistical Risk Assessment: Old Problems and New Applications, 52 Crime & Delinq. 178, 180 (2006). The Board members then make a recommendation based on this brief interview. See id. This antiquated assessment method neither adequately assesses the risk of recidivism nor creates a credible, transparent system that is fair to inmates, practitioners, and judges who review the record. Subjective assessment also negatively impacts inmates and society by keeping low-risk inmates in jail, increasing their risk of recidivism and expanding the already outsized prison population. A. The Board’s current subjective assessment method fails to adequately consider and reduce recidivism. The Board’s subjective assessment does not assess recidivism effectively because it fails to consider relevant information or analyze it in a systematically 6 rational method. Risk of future criminality is difficult to predict, and researchers have found that a complex and interwoven set of static and dynamic factors influences recidivism. See Joseph T. McCann, Risk Assessment and the Prediction of Violent Behavior, Federal Lawyer, Oct. 1997, at 18–19 (discussing factors contributing to recidivism). While static factors can predict recidivism—factors that the inmate cannot change, such as prior criminal history and the crime of conviction—so do dynamic factors—factors that the inmate can affect while incarcerated and improve with the assistance of family and community, including institutional record, programming, education, good behavior, and remorse. See Susan Turner & Julie Gerlinger, Risk Assessment and Realignment, 53 Santa Clara L. Rev. 1039, 1044 (2013). The current assessment method ignores much of this information. Edward R. Hammock & James F. Seelandt, New York’s Sentencing and Parole Law: An Unanticipated and Unacceptable Distortion of the Parole Boards’ Discretion, 13 St. John’s J. Legal Comment. 527, 550 (1999). Moreover, recidivism risk is much too complex for three Board members to assess in a quick interview, especially when untrained in criminology or psychology. See Tim Brennan, William Dieterich & Beate Ehret, Evaluating the Predictive Validity of the Compas Risk and Needs Assessment System, 36 Crim. Just. & Behav. 21, 21 (2009). 7 Because Board members neither weigh the full array of factors influencing recidivism risk, nor do so in a rationalized, systematic fashion, they often rely on their own subjective sense of justice. Even though “institutional record is one of the ‘most significant factors’ in predicting [an inmate’s] ability to assume his place in a free society,” the Board instead usually focuses on the nature of past crimes. Shepard v. Taylor, 556 F.2d 648, 652 (2d Cir. 1977) (quoting Moody v. Daggett, 429 U.S. 78, 88–90 (1976)). Time after time, the Board denies parole after focusing almost exclusively on static factors and offering subjective conjecture about recidivism risk. See King v. New York State Div. of Parole, 190 A.D.2d 423, 432–33 (1st Dep’t. 1994) (“[T]he record also reveals a failure on the Board’s part to consider and fairly weigh all of the information available to them concerning petitioner that was relevant. . . .”); Rabenbauer v. New York State Dep’t of Corr. & Cmty. Supervision, 46 Misc.3d 603, 608–09 (N.Y. Sup. Ct. Sullivan County 2014) (holding that “[t]here is no additional rationale, other than the board’s opinion of the heinous nature of the instant offense, and personal beliefs and speculations, to justify the denial of parole release”); Coaxum v. New York State Bd. of Parole, 827 N.Y.S.2d 489, 494–95 (N.Y. Sup. Ct. Bronx County 2006) (“Here, however, the Board’s decision reveals it accorded no weight and no emphasis whatsoever to any factor apart from the seriousness of petitioner’s offense”) (emphasis in original); Rios v. New York State Div. of Parole, 15 Misc.3d 1107A, at *4 (N.Y. Sup. Ct. 8 Kings County 2007) (“[T]he Parole Board focused almost exclusively on the serious nature of petitioner’s crime.”) In short, the Board’s subjective assessment method—and the resulting narrow focus on past crimes, often decades past—does not appropriately address recidivism risk. See D.A. Andrews, James Bonta, & J. Stephen Wormith, The Recent Past and Near Future of Risk and/or Need Assessment, 52 Crime & Delinq. 7, 13 (2006). The Board needs a scientific, integrated, and holistic approach, such as COMPAS, to properly evaluate all the factors and evidence relevant to recidivism risk using a rationally structured method. B. The subjective assessment method is susceptible to errors and opaque. The Board’s current subjective assessment is prone to error and lacks transparency. Its members’ intuitive judgments are susceptible to such biases as representativeness (believing one instance represents the entire population) and availability (overvaluing vivid data). See William M. Grove et al., Clinical Versus Mechanical Prediction: A Meta-Analysis, 12 Psychol. Assessment 19, 25 (2000). These likely lead Board members to both overvalue how the crime was committed because it is particularly vivid and to extrapolate from that crime to other aspects of the inmate. Other common errors include ignoring base rates, assigning non- optimal weights to cues, failure to take into account regression toward the mean, and failure to properly assess covariation. Id. These errors result in biased 9 decision making. See Brennan et al., supra, at 21. Importantly, Board members do not receive feedback on the accuracy of their predictions; after an inmate is released or denied parole, Board members are unlikely to follow up. See Grove et al., supra, at 25. As a result, Board members lack the information necessary to identify patterns of error or to improve their assessments over time. Furthermore, the Board members’ failure to memorialize their reasoning when denying parole conceals these errors. “The typical parole release denial often follows a predictably textbook form and employs . . . trademark-boilerplate language.” See Hammock & Seelandt, supra, at 535. This leads to “decision rules not easily observable and . . . difficult to replicate.” James Bonta, Risk-Needs Assessment and Treatment, in Choosing Correctional Options That Work: Defining the Demand and Evaluating the Supply, 18, 19 (A.T. Harland ed. 1996). The lack of transparency results in three systemic problems. First, “because the decision lacks any support from the record,” judges are “left with no ability to evaluate why the commissioners on the board denied parole.” See Hawkins v. New York State Dep’t of Corr. & Cmty. Supervision, Sup. Ct. Sullivan County, May 11, 2015, at *8 (Index No. 0011-15). See also Canales v. Hammock, 105 Misc. 2d 71, 74 (N.Y. Sup. Ct. 1980) (“If the board chose not to believe the statements of petitioner at the hearing, then it should have detailed its reasons for discrediting his opposing statement in its decision.”) (emphasis in original). Second, inmates are 10 completely unable “to figure out what it is [they] should do to convince any parole board that [they are] rehabilitated and ready for release back into society.” See Hawkins, at *8. “One can imagine nothing more cruel, inhuman and frustrating than serving a prison term without knowledge of what will be measured and the rules determining whether one is ready for release.” Franklin v. Shields, 569 F.2d 784, 792 (4th Cir. 1977) (citation omitted). Finally, using boilerplate language when denying parole allows Board members to make decisions without reflecting why they denied parole. This lack of self-reflection perpetuates biases and pre- determined judgments. The cost of these errors is incalculable. Denying parole when inmates should be released harms the inmate, the prison system, and the family and social fabric. For low-risk inmates, staying in prison increases recidivism risk because the inmates make criminal contacts and are increasingly isolated from support outside prison. See Christopher Lowenkamp & Edward Latessa, Understanding the Risk Principle: How and Why Correctional Interventions Can Harm Low-Risk Offenders, Topics in Community Corrections, at 7 (2004) (finding that more incarceration for low-risk individuals increases recidivism because of high-risk inmates’ negative influence). For prisons, keeping those inmates who should be released only adds to the substantial and expensive incarcerated population. 11 Lastly, denying parole to inmates who should be released disrupts families and the communities where they belong. The effects of parental incarceration on the family, which has been likened to bereavement, is well documented. See Travis, J., Western, B., Redburn, S. (Eds.)(2014). The Growth of Incarceration in the United States: Exploring Causes and Consequences. Washington, D.C.: The National Academies Press; Christopher Uggen & Suzy McElrath, Parental Incarceration: What We Know and Where We Need to Go, 104 J. Crim. L. & Criminology 597, 598 (2014). Likewise pronounced is the effect of longer sentences on an inmate’s ability to readjust and contribute to society upon release. See Lin Song & Roxanne Lieb, Recidivism: The Effect of Incarceration and Length of Time Served, Wash. State Inst. Pub. Pol’y (1993); see generally Mary E. Pattillo et al., Imprisoning America: the social effects of mass incarceration (Russell Sage Foundation) (2004); Megan Comfort, Doing time together: love and family in the shadow of the prison (University of Chicago Press) (2008). Keeping in jail a prisoner who should be released exacerbates these baleful effects. 12 II. NON-DISCRETIONARY USE OF COMPAS WOULD SUBSTANTIALLY IMPROVE NEW YORK STATE’S PAROLE DETERMINATIONS. COMPAS is a fourth-generation3 risk and needs instrument that substantially improves the Parole Board’s capacity to determine whether an inmate is likely to reoffend. It was created from over 30,000 assessments conducted between January 2004 and November 2005 across the United States, measuring intermediate outcomes and relating them to recidivism risk. See Northpointe, Practitioners Guide to COMPAS, at 2 (2012). COMPAS is a proven tool that can create better treatment plans than past generations of risk and needs instruments because it takes a holistic approach focused on rehabilitation efforts. See Andrews, Bonta & Wormith, supra, at 8. In so doing, COMPAS can reduce the errors, and therefore the costs, of the subjective assessments currently in practice. A. How COMPAS works COMPAS is based on a broad range of forward-looking factors. As the first step of a COMPAS assessment, a parole officer and inmate answer a questionnaire containing almost 100 questions that relate to family, criminal associates, education, and financial status. The answers are analyzed by the COMPAS 3 The first generation of risk assessment consisted of unstructured professional judgments. Board members essentially guessed based on their experience and an interview whether someone should be released. Second generation instruments introduced some empirics, but the inputs were atheoretical and static. The third generation improved the empirics by adding a wider range of factors that were dynamic and theoretically informed. Finally, fourth generation models followed inmates from intake through case closure. 13 software, yielding a recidivism risk score. COMPAS estimates these scores in deciles, where 1 is at the low end and 10 at the high end. If an inmate receives a 3, that inmate is in the thirtieth percentile: 30 percent of people are less likely to commit a crime if released and 70 percent are more likely. The decile scores are partitioned into three risk levels: low, medium, and high. Low-risk offenders score 1–4, the lowest 40% of cases; medium-risk offenders score 5–7; and high-risk offenders score 8–10. COMPAS also produces a decile score for each “criminogenic need,” such as criminal associates, substance abuse, and vocational problems. Each criminogenic need is correlated with recidivism—as the decile score for the need increases, so does the inmate’s recidivism risk. Board members receive 22 decile scores: three overall risk potential scores (violence, general recidivism and failure to appear) and nineteen criminogenic need scores.4 These scores both help parole officers determine when an inmate is ready for release and help them tailor a treatment plan after release. In turn, officers can direct intensive services to high-risk offenders, reduce services to low-risk offenders, and provide treatment in a mode responsive to the inmate’s learning style and ability. See D.A. 4 The criminogenic needs are: criminal involvement, history of non-compliance, history of violence, criminal attitude, resentful/mistrust, responsivity problems, few pro-social peers, criminal associates, impulsivity, anger, family supports, family criminality, substance abuse, life goals, financial problems, educational problems, social isolation, social environment and depression. 14 Andrews & James Bonta, Rehabilitating Criminal Justice Policy and Practice, 16 Psychol. Pub. Pol’y L. 39, 44–45 (2010). To produce the decile scores, COMPAS is first customized and calibrated to its target population within the jurisdiction where it is being applied. See id. (“COMPAS core assessment is designed to be configurable for use … with various populations.”). Therefore, in New York, COMPAS would be customized to address the characteristics of the state’s parole population. COMPAS also uses advanced statistical modeling similar to that used in setting car insurance rates. The inputs are static and dynamic factors that correlate with recidivism. Static factors are such immutable characteristics as criminal history and history of drug abuse. Dynamic factors can change, and include employment status, criminal associates, achievements, and education. While static factors are useful predictors of recidivism, incorporating dynamic factors improves upon past recidivism risk models. “Tools that contain dynamic risk factors are considered more predictive than tools that contain static factors that are immutable.” Turner & Gerlinger, supra, at 1044. Moreover, dynamic factors help Board members tailor their approach to parolees effectively to reduce recidivism because officers can target specific needs identified by the dynamic factors. See Andrews & Bonta, supra, at 46; James Bonta, Offender Risk Assessment: Guidelines for Selection and Use, 29 Crim. Just. & Behav. 355, 370 15 (2002) (“To manage risk effectively, knowledge of dynamic risk factors is critical.”). For example, if recidivism risk increases because an inmate has poor employment prospects, parole officers can work with the inmate to reduce that risk through employment counseling. Finally, dynamic factors measure the inmate’s change during incarceration. Measuring change paints a clearer picture of when a person is ready to reintegrate into society. It also motivates rehabilitation in offenders because they know rehabilitation in prison matters in parole decisions. COMPAS and the decile scores for recidivism risk—even when their use is not discretionary—do not replace the Board. Instead, COMPAS provides more evidence to facilitate reasoned and clear decisions by the Board. The developers of COMPAS also candidly expect disagreement with the instrument “in about 10% of the cases due to mitigating or aggravating circumstances which the computer is not sensitive to.” Practitioners Guide to COMPAS, supra, at 31. In cases of disagreement, Board members can (and should) document their reasons for deviating from COMPAS. See id. This allows other evaluators and the inmate to understand the decision process. B. Other states have succeeded with risk and needs assessment tools. Many states (as many as 24, according to a 50-state survey published in 2012) have utilized such risk and needs assessment instruments as COMPAS in their parole system. See Amy Robinson-Oost, Note, Evaluation as the Proper 16 Function of the Parole Board: An Analysis of New York State’s Proposed Safe Parole Act, 16 CUNY L. Rev. 129, 144 (2012). And many have combined reliance on these instruments with some opportunity for structured independent judgment by the parole board. See, e.g., Colo. Rev. Stat. Ann. § 17-22.5-404 (West) (stating that risk assessment tools can predict the likelihood or risk of reoffense with significantly greater accuracy than professional judgment alone and that the best outcomes are derived from a combination of empirically based tools and clinical judgment). States implementing risk and needs instruments have seen positive results. A recent report found that these instruments reduced recidivism in nine states. See Nat’l Reentry Res. Ctr., Reducing Recidivism: States Deliver Results 4 (2014). For example, North Carolina’s recidivism rate declined by 19.3 percent from 2007 to 2013 in part because the state emphasized individual case planning using a risk and needs tool. See id. at 9. Kansas reduced the number of parole revocations by 25 percent between 2006 and 2009 after incorporating a risk and needs tool. See Christine S. Scott-Hayward, The Failure of Parole: Rethinking the Role of the State in Reentry, 41 N.M. L. Rev. 421, 456–57 (2011). Similarly, Texas decreased its parole revocation rate by 25 percent between 2006 and 2008 after implementing evidence-based tools. See id. Even if reducing the parole revocation rate were not the goal, states have found that “[s]tructured decision- 17 making by the state board of parole provides for greater accountability, standards for evaluating outcomes, and transparency of decision-making that can be better communicated to victims, offenders, other criminal justice professionals, and the community.” Colo. Rev. Stat. Ann. § 17-22.5-404 (West). C. COMPAS outperforms the subjective assessment method. Non-discretionary use of COMPAS, a scientific and statistically validated instrument, makes five critical improvements on the status quo. First, COMPAS more accurately predicts recidivism than subjective assessment. Second, it increases consistency across parole decisions because all Board members, by using COMPAS, will make decisions using a standardized framework based on similar amounts and quality of information. Third and relatedly, COMPAS eliminates the bias inherent in the subjective assessment method. Fourth, COMPAS provides Board members a broader and more robust evidence base. Finally, COMPAS adds transparency to parole decisions. i. COMPAS better predicts likelihood of recidivism than subjective assessment. Empirical studies consistently demonstrate that COMPAS accurately predicts recidivism. See Brennan et al., supra, at 21 (2009); Division of Criminal Justice Services, New York State COMPAS-Probation Risk and Need Assessment Study: Examining the Recidivism Scale’s Effectiveness and Predictive Accuracy, at 6 (2012); Bryn A. Herrschaft, Evaluating the Reliability and Validity of the 18 Correctional Offender Management Profiling for Alternative Sanctions (COMPAS) Tool: Implications for Community Corrections Policy, at 98 (Jan. 2014) (unpublished Ph.D. dissertation, Rutgers University). In New York, COMPAS accurately predicts whether a person will commit a crime again 71% of the time. See New York State COMPAS-Probation Risk and Need Assessment Study, supra, at 18. Studies also demonstrate that evidence-based risk assessments, such as COMPAS, predict recidivism better than subjective assessment methods alone. See New York State Comm’n on Sentencing Reform, The Future of Sentencing in New York State: Recommendations for Reform, supra, at 156; Gottfredson & Moriarty, supra, at 180 (“In virtually all decision-making situations, it has been found that [empirically] developed predictions outperform human judgments”); Bonta, supra, at 357 (“There is ample evidence that [empirical] assessments of risk are significantly superior to [subjective] assessments”); William M. Grove & P.E. Meehl, Comparative Efficiency of Informal (Subjective, Impressionistic) and Formal (Mechanical, Algorithmic) Prediction Procedures: the Clinical-Statistical Controversy, 2 Psychol. Pub. Pol’y L. 293 (1996) (suggesting that not using empirical instruments may be unethical and unprofessional because empirical instruments better predict recidivism). By contrast, meta-analysis—a study that analyzes all studies in a particular field—finds that subjective assessment of 19 inmates poorly predicts outcomes. See Andrews, Bonta & Wormith, supra, at 12. Therefore, non-discretionary use of COMPAS will improve the reliability and accuracy of recidivism predictions in the New York parole system. ii. COMPAS increases consistency in parole decision making. Non-discretionary use of COMPAS increases consistency because it relies on objective criteria applicable to every parole-eligible inmate. By contrast, the subjective assessment method treats each inmate differently because the method depends on Board members’ intuitions. Moreover, Board members do not always recognize many similarities between inmates because of their tendency to narrowly focus on how the crime was committed. COMPAS, by contrast, helps develop an individualized recidivism risk profile for each inmate, but one based on uniform inputs and method. Thus, similar inmates will have similar risk profiles, reducing inconsistent decisions amongst Board members. iii. COMPAS helps to eliminate implicit bias. Non-discretionary use of COMPAS will reduce bias because it will cause Board members to evaluate and document why they reach a conclusion different from one recommended by the risk assessment instrument. If a Board member’s decision conflicts with the risk assessment, that Board member is placed in the position of articulating valid reasons for overriding the tool. Alternatively, after observing the non-discretionary COMPAS result, the Board member may self- 20 evaluate the reasons why his or her subjective assessment may reach different conclusions and critically examine that assessment. Over time, this self-check and feedback mechanism will foster better decision making. iv. Decisions made using COMPAS rest on a more robust evidence base. Decisions made using COMPAS incorporate more robust evidence because COMPAS reflects far more assessments of New York inmates than Board members could consider and systematizes the information in a manner according to statistical analysis in a way members could not be expected to do. Furthermore, because the tool is specifically configured to New York and its parole population, COMPAS better reflects the environmental and socioeconomic pressures facing potential parolees. Therefore, non-discretionary use of COMPAS systematically accounts for evidence that an individual Board member could not. In addition, the questions asked in the COMPAS questionnaire are broader than those asked during a parole interview, which usually focuses on how the inmate committed the crime. This information also shows who the inmate is today and how the inmate will reintegrate after release. Consistently focusing on today and taking a holistic approach better predicts how the inmate will perform when released. v. COMPAS promotes transparency. Non-discretionary use of COMPAS promotes transparency on an individual and systemic level. It produces intelligible records that can be shared with inmates 21 and that in turn can guide and aid the inmates in their “endeavor to return to society as a useful citizen.” See Greene v. Smith, 52 A.D.2d 292, 294, 384 N.Y.S.2d 901 (4th Dep’t 1976). Non-discretionary use of COMPAS also promotes transparency on a systemic level by facilitating meaningful appellate review. CONCLUSION The value of COMPAS diminishes when it is used in an ad hoc, discretionary manner to supplement a Parole Board member’s subjective assessment of a candidate’s risk of recidivism. None of the process values reflected in the 2011 statutory amendments—accuracy, credibility, transparency—are advanced in a parole system that fails to establish written procedures to use scientifically valid tools in a consistent, structured manner. COMPAS must at a minimum, therefore, be established by procedures as a uniform framework and not merely a discretionary factor in assessing whether an individual is suitable for parole. Accordingly, amici respectfully urge the Court to modify the order of the Appellate Division.