An actuarial model for assessment of prison violence risk among maximum security inmates

An experimental scale for the assessment of prison violence risk among maximum security inmates was developed from a logistic regression analysis involving inmates serving parole-eligible terms of varying length (n = 1,503), life-without-parole inmates (n = 960), and death-sentenced inmates who were mainstreamed into the general prison population (n = 132). Records of institutional violent misconduct of these 2,595 inmates were retrospectively examined for an 11-year period (1991 to 2002). Predictors affecting the likelihood of such misconduct included age, type and length of sentence, education, prior prison terms, prior probated sentences, and years served. The scale was modestly successful, as demonstrated by an overall omnibus area under the curve of .719. Double cross-validation demonstrated minimal shrinkage. The authors have termed this experimental scale the Risk Assessment Scale for Prison.     

Actuarial models for assessing prison violence risk: revisions and extensions of the Risk Assessment Scale for Prison (RASP)

An investigation and extension of the Risk Assessment Scale for Prison (RASP-Potosi), an actuarially derived scale for the assessment of prison violence, was undertaken through a retrospective review of the disciplinary records of the first 12 months of confinement of a cohort of inmates entering the Florida Department of Corrections in 2002 and remaining throughout 2003 (N=14,088). A near replication of the RASP-Potosi and additional analyses based on other weighted logistic regression models were performed on an inmate subsample for whom all information categories were available (n=13,341). Younger age and shorter sentences were associated with increased violent misconduct. Older age, drug conviction, and higher educational attainment were associated with reduced violent misconduct. Regardless of whether the original RASP-Potosi or its progeny were utilized, or the custody level of the inmate sample, the models were modestly successful in predicting prison violence, with the area under the curve (AUC) ranging from .645 to .707.     

In search of the “Tossed Salad Man” (and others involved in prison violence): New strategies for predicting and controlling violence in prison

This article addresses two fundamental questions: 1) Can we accurately identify a subgroup of high risk inmates who will likely be involved in various forms of prison violence; and if so, 2) Is there an empirical research supporting the contention that current classification systems reduce the risk posed by those offenders predicted to be violent (such as gang/security threat group members)? Based on a review of the empirical evidence from the past two decades, we conclude that current classification strategies do not predict prison violence very accurately and perhaps more importantly, they do not appear to reduce the risk of violence in prison. New classification systems focusing on changing – rather than controlling – offenders while in prison represent one possible alternative strategy, given recent evidence that participation in prison programming/treatment is the most effective prison violence reduction strategy currently available.     

情感计算技术如何推动犯罪风险评估工具的发展？

摘 要 随着近年来人工智能深度学习技术的发展,情感计算与人格计算技术日渐成熟,在许多实际应用场景中取得了良好的表现,当前人工智能情感计算技术应用于犯罪风险评估领域,能够有效解决目前主流的风险评估工具难以解决的个体内差异性的预测因子评估问题以及被测评参与者因社会赞许性而导致结果失真的问题。本文在详细阐述目前主流评估工具的局限性基础上,详细阐述了以情感计算技术为支撑的动态风险评估工具的设计思路、目前已有的技术方案以及设计细节的理论依据,在此基础上最后提出以人工智能技术为支撑的新型评估工具的未来发展方向。     

An empirical typology of violent incidents reported by prison inmates with convictions for violence

Forty-four male prison inmates convicted for violent offences were interviewed, and details of current offence, previous convictions for violence, prison incidents, and violent situations outside of prison were obtained from each offender. Information on victim, circumstances, precipitating events, type of violence, emotions at the time, and the like, was derived for 246 violent incidents. Cluster analysis revealed eight clusters: “criminal” violence to familiar victims, peer violence within prison, domestic violence, peer violence in pubs, gang violence, violence to staff in prison, “criminal” violence to unknown victims with accomplices, and public violence against younger victims. Results were discussed in terms of treatment of violent offenders, with special emphasis on Social Skills Training.     

Applicability of two violence risk assessment tools in a psychiatric prison hospital population

The risk of violent behavior is known to be higher for patients who suffer from a severe mental disorder. However, specific prediction tools for clinical work in prison psychiatry are lacking. In this single-center study, two violence risk assessment tools (Forensic Psychiatry and Violence Tool, “FoVOx,” and Mental Illness and Violence Tool, “OxMIV”) were applied to a prison hospital population with a primary psychotic or bipolar disorder and subsequently compared. The required information on all items of both tools was obtained retrospectively for a total of 339 patients by evaluation of available patient files. We obtained the median and inter-quartile range for both FoVOx and OxMIV, and their rank correlation coefficient along with 95% confidence intervals (CIs)—for the full cohort, as well as for cohort subgroups. The two risk assessment tools were strongly positively correlated (Spearman correlation = 0.83; 95% CI = 0.80–0.86). Such a high correlation was independent of nationality, country of origin, type of detention, schizophrenia-spectrum disorder, previous violent crime and alcohol use disorder, where correlations were above 0.8. A lower correlation was seen with patients who were 30 years old or more, married, with affective disorder and with self-harm behavior, and also in patients without aggressive behavior and without drug use disorder. Both risk assessment tools are applicable as an adjunct to clinical decision making in prison psychiatry.     

Predicting future reconviction in offenders with intellectual disabilities: the predictive efficacy of VRAG, PCL-SV, and the HCR-20

Accurate predictions of future reconviction, including those for violent crimes, have been shown to be greatly aided by the use of formal risk assessment instruments. However, it is unclear as to whether these instruments would also be predictive in a sample of offenders with intellectual disabilities. In this study, the authors have shown that the Violence Risk Appraisal Guide (V. L. Quinsey, G. T. Harris, M. E. Rice, & C. Cormier, 1998); the Psychopathy Checklist--Screening Version (S. D. Hart, D. N. Cox, & R. D. Hare, 1995); and the History, Clinical, Risk Management--20 (C. D. Webster, K. S. Douglas, D. Eaves, & S. D. Hart, 1997) were all significant predictors of violent and general reconviction in this sample, and in many cases, their efficacy was greater than in a control sample of mentally disordered offenders without an intellectual disability.     

Zur Validit?t des Prognoseinstruments ?Rückfallrisiko bei Sexualstraft?tern“

The Risk for Recidivism in Sexual Offenders (RRS) procedure is an actuarial risk assessment tool designed to predict recidivism in sexual offenders released from prison. So far it is the only tool developed with a German sample and uses a combination of 11 static and dynamic risk markers. The goal of the current study was to test the predictive validity of the 3 RRS prognostic scores in a sample of 104 sexual offenders whose prison files and recidivism data were analysed after a mean period of 8 years after release from a German prison. The H-score of the RRS developed to assess the probability of reimprisonment showed moderate to good predictive power in relation to any new convictions, new convictions leading to a prison term and new convictions for an aggressive or sexual criminal offence. As expected it was not able to predict sexual recidivism which had a base rate of 8.7% in this sample. The S-score and the K-score of the RRS developed to predict this criterion, performed hardly better in this task. These scores showed moderate to good predictive validity only in the small subsample of extrafamilial sexual offenders.     

Die Vorhersage der Rückfälligkeit entlassener Sexualstraftäter mittels des Sexual Offender Risk Appraisal Guide (SORAG) und dessen Screening-Version (SORAG-SV): Darstellung der differentiellen und prädiktiven Validität

The Sex Offender Risk Appraisal Guide (SORAG) is one of the most important actuarial risk assessment instruments for sexual offenders. In the meantime, the application of actuarial instruments in recidivism risk assessment for sexual offenders is regarded as state of the art. In addition to results about interrater reliability and concurrent validity we present results about the differential and predictive validity of the SORAG and a screening version of the SORAG (SORAG-SV), which relies only on file information. In order to examine the predictive validity, we used a representative sample of 519 male sexual offenders released from a prison sentence served in one of Austrian´s prisons with a mean follow-up period of 3½ years. Furthermore, we tested the differential validity by dividing the whole sample in different subsamples regarding age, index offense type, and degree of antisociality. Both SORAG and SORAG-SV showed predominantly good predictive accuracy which, however, varied depending on offender subgroup and recidivism category.     

Assessment of accused juvenile sex offenders in Germany: a comparison of five different measures

The present study examines the relationship between risk and protective factors among young alleged sexual offenders (N = 66) in pre-trial and pre-treatment settings. For risk assessment purposes, the Screening Tool for the Assessment of Young Sexual Offenders' Risk (STAYSOR), the Structured Assessment of Violence Risk in Youth (SAVRY), and the Structured Assessment of Protective Factors for violence risk (SAPROF) were used. Psychopathological indicators measured with the Basis Raads Onderzoek (BARO) and the German adaption of the Reynolds Adolescent Adjustment Screening Inventory (RAASI) were used to examine the relationship between the risk assessment scales and instruments measuring risk-relevant psychopathological constructs. Risk and protective factors were significantly negatively correlated. Psychopathological measures were positively correlated with risk factors and negatively with protective factors. Although further studies on the predictive validity of the instruments are needed, the results of the present explorative pilot study indicate that the use of all five instruments may be clinically meaningful for the assessment of young persons who are at risk of sexual offending.     

A comparative study of adolescent risk assessment instruments: predictive and incremental validity

Promising new adolescent risk assessment tools are being incorporated into clinical practice but currently possess limited evidence of predictive validity regarding their individual and/or combined use in risk assessments. The current study compares three structured adolescent risk instruments, Youth Level of Service/Case Management Inventory (YLS/CMI), Structured Assessment of Violence Risk in Youth (SAVRY), and Psychopathy Checklist: Youth Version (PCL:YV), for both predictive and incremental validity with respect to general and violent recidivism. Receiver operating characteristic and hierarchical logistic regression analyses revealed that the risk tools predicted general and violent recidivism to varying degrees of accuracy, but the SAVRY offered the most in incremental validity. Clinical implications and future directions for youth risk assessment are discussed.     

Verläufe bei sexuellen Tötungen

This article summarizes main results of studies on forensic psychiatric court reports on 166 men who had been persecuted between 1963 and 1991 for a sexual offence leading to the death of the victim. Comparing perpetrators with a single victim and those with multiple victims we found similar results as in two previous studies with smaller samples: Multiple sexual homicide perpetrators showed more often sexual sadism and other paraphilias, as well as antisocial, schizoid and sadistic personality disorders. Follow-up data from the federal criminal records could be obtained for 139 offenders. Ninety perpetrators had been released after a mean detention of 12.2 years, whereas the 49 offenders who were still in prison or forensic psychiatric hospitals had been detained for a mean period of 20.6 years. The non-released offenders showed more often paraphilias as well as antisocial and sadistic personality disorders than the released perpetrators. Paraphilias and antisocial personality traits are empirically well proven risk factors for criminal recidivism with sexual reoffences. In addition, the non-released sexual homicide perpetrators had higher scores in all applied risk assessment instruments (PCL-R, HCR-20, SVR-20, Static-99). Among the released offenders only 1.1% (n=1) reoffended with a completed homicide and 2.2% (n=2) with attempted homicide. The recidivism rates with sexual and other violent reoffences in this sample of sexual homicide perpetrators were similar to those in a large meta-analysis on recidivism in sexual offenders by Hanson and Morton-Bourgon (4). Since well established risk factors had apparently been “used-up” for the decisions about release or non-release, in the follow-up data about the released offenders only age at the sexual homicide and age at the time of release were found as risk factors for recidivism with any violent (sexual or non-sexual) reoffence, i.e. the younger the offender at the time of the homicide and the younger at the time of release, the more likely is the risk of violent reoffending.     

Men who molest their sexually immature daughters: is a special explanation required?

Child molesters who target their own children have been described as low risk and not pedophilic. Men who had molested a daughter or stepdaughter (n = 82) were compared to 102 molesters whose only female victims were extrafamilial. Men who offended against their own daughters had less deviant sexual age preferences and were less likely to commit new violent and sexual offenses. However, the father-daughter molesters exhibited an average absolute phallometric preference for prepubertal children and had a violent recidivism rate of 22% in a follow-up of less than 5 years. Actuarial risk assessment instruments (the Violence Risk Appraisal Guide and the Sex Offender Risk Appraisal Guide; V. L. Quinsey, G. T. Harris, M. E. Rice, and C. A. Cormier, 1998) worked as well for intrafamilial child molesters as for other sex offenders.     

Über die Treffsicherheit der standardisierten Risikoeinschätzungsverfahren Static-99 und Stable-2007 bei aus einer Sicherungsmaßnahme entlassenen Sexualstraftätern

The accuracy of the recidivism risk assessment instruments Static-99 and Stable-2007 for sexual offenders was examined in a population of released male forensic sexual offenders hospitalized under mandatory treatment in Austria (N?=?96). The Static-99 with an area under the curve (AUC) ?value of 0.86 and the Stable-2007 (AUC?=?0.71) were significantly related to sexual reoffending after nearly 7 years time at risk, thus revealing a predictive power comparable with offenders released from prison (N?=?274). Also the Stable-2007 incrementally supplemented the predictive accuracy of the Static-99. Static-99/Stable-2007 risk/need categories identified a high risk group with a 50?% chance for sexual reconviction within 5 years after release despite a favorable risk assessment as the precondition for release and post-release risk management. On the other hand, there was virtually no relapse in the three lowest risk categories suggesting a specific effect of the mandatory treatment at least in these offender categories. The data suggest that the instruments are valid not only for offenders released from prison but also for forensic sexual offenders.     

Violence in criminal careers: A review of the literature from a developmental life-course perspective

The criminal career paradigm focuses on the prevalence and frequency of crime along with an emphasis on offense specialization, age of onset and desistance, offense seriousness, and career length. Coinciding with the criminal career framework, developmental/life-course criminology offers more complex theoretical explanations for crime which highlight the importance of age-graded developmental risk factors that affect offending over the course of a criminal career. This review provides a systematic examination of the role of violence in criminal career and developmental/life-course research with a specific focus on its prevalence and frequency and its share of the offense repertoire. The results show that the incidence of violence is rare in one's criminal career except for a small group of chronic offenders who are responsible for a majority of the violent offenses. Regarding specialization, evidence suggests that offenders are spectacularly non-specialized and violent offenders can primarily be characterized as frequent offenders who offend more often and thereby have a higher probability of committing a violent offense in their criminal career. Conclusions and directions for future research are discussed.     

A dynamic assessment of offender risk, needs, and strengths in a sample of pre-release general offenders

Static, dynamic, and protective factors have been identified as three focal domains significantly related to offender recidivism. However, few measures include comprehensive and inclusive assessment of these variables. The Inventory of Offender Risk, Needs, and Strengths (IORNS) was developed to fill a void in the assessment of risk and needs for offenders. The current study examines the reliability and initial validity of the IORNS in a sample of pre-release offenders assessed for risk and treatment need. Results indicate moderate to high levels of internal consistency and identical IORNS scale results across race, and that the IORNS indexes, scales, and subscales display good convergent validity with self-report and interview measures of static risk, dynamic risk, antisocial behavior, psychopathy, personality pathology, substance abuse, depression, and anxiety. Initial predictive validity examination of the IORNS indicates that several of the indexes, scales, and subscales were able to differentiate offenders who were sent back to prison for half-way house rule violations from those who did not violate rules.     

Probability of Criminal Acts of Violence: A Test of Jury Predictive Accuracy

The ability of capital juries to accurately predict future prison violence at the sentencing phase of aggravated murder trials was examined through retrospective review of the disciplinary records of 115 male inmates sentenced to either life (n = 65) or death (n = 50) in Oregon from 1985 through 2008, with a mean post‐conviction time at risk of 15.3 years. Violent prison behavior was completely unrelated to predictions made by capital jurors, with bidirectional accuracy simply reflecting the base rate of assaultive misconduct in the group. Rejection of the special issue predicting future violence enjoyed 90% accuracy. Conversely, predictions that future violence was probable had 90% error rates. More than 90% of the assaultive rule violations committed by these offenders resulted in no harm or only minor injuries. Copyright © 2013 John Wiley & Sons, Ltd.     

Risk factors in violent and nonviolent offenders

Factors leading to risk for violence were evaluated in an offender population of 36 male offenders including 18 Violent, e.g., assault, threatening and 17 nonviolent (break and enter, theft), and 17 nonoffenders. Their scores on the Psychopathy Checklist Revised, Violence Risk Scale-Experimental Version 1, Minnesota Multiphasic Personality Inventory-168, and the Porteus Maze tests showed scores for the inmates with violent offenses were elevated on Psychopathic Deviate, Paranoia, Schizophrenia scales of the Minnesota Multiphasic Personality Inventory, violence risk of the Violence Risk Scale, showed psychopathic orientation on the Psychopathy Checklist Revised, and had a lower test age quotient score on the Porteus Maze test.     

Violent offenses associated with co‐occurring substance use and mental health problems: Evidence from CJDATS

The present study examines the relationship between substance use, mental health problems, and violence in a sample of offenders released from prison and referred to substance abuse treatment programs. Data from 34 sites (n = 1,349) in a federally funded cooperative, the Criminal Justice Drug Abuse Treatment Studies (CJDATS), were analyzed. Among parolees referred to substance abuse treatment, self‐reports for the six‐month period before the arrest resulting in their incarceration revealed frequent problems with both substance use and mental health. For most offenders with substance use problems, the quantity of alcohol consumed and the frequency of drug use were associated with a greater probability of self‐reported violence. Mental health problems were not indicative of increases in violent behavior, with the exception of antisocial personality problems, which were associated with violence. The paper emphasizes the importance of providing substance abuse treatment in relation to violent behavior among offenders with mental health problems being discharged to the community. Copyright © 2009 John Wiley & Sons, Ltd.