Machine Learning and Risk Assessment: Random Forest Does Not Outperform Logistic Regression in the Prediction of Sexual Recidivism.

Although many studies supported the use of actuarial risk assessment instruments (ARAIs) because they outperformed unstructured judgments, it remains an ongoing challenge to seek potentials for improvement of their predictive performance. Machine learning (ML) algorithms, like random forests, are able to detect patterns in data useful for prediction purposes without explicitly programming them (e.g., by considering nonlinear effects between risk factors and the criterion). Therefore, the current study aims to compare conventional logistic regression analyses with the random forest algorithm on a sample of N = 511 adult male individuals convicted of sexual offenses. Data were collected at the Federal Evaluation Center for Violent and Sexual Offenders in Austria within a prospective-longitudinal research design and participants were followed-up for an average of M = 8.2 years. The Static-99, containing static risk factors, and the Stable-2007, containing stable dynamic risk factors, were included as predictors. The results demonstrated no superior predictive performance of the random forest compared with logistic regression; furthermore, methods of interpretable ML did not point to any robust nonlinear effects. Altogether, results supported the statistical use of logistic regression for the development and clinical application of ARAIs.

A Moderated Mediation Analysis to Further Examine the Role of Verbal Intelligence in the Association Between Psychopathic Personality and Crime.

The current study examined the association between psychopathy, criminal behavior, and the role of verbal intelligence. One promising approach is to examine alternative links between psychopathic traits and criminality like moderation and mediation effects by considering the potential relevance of verbal intelligence as a possible moderating variable. We hypothesized that psychopathic traits linearly predict antisocial behavior (ASB) but that a conviction because of ASB is moderated by verbal intelligence. To test a path model of this hypothesis, N = 305 participants (42% women; n = 172 inmates of German correctional facilities) filled in questionnaires to assess psychopathic traits, ASB, criminal behavior, and verbal intelligence. The moderated mediation analysis revealed that high psychopathic traits go along with a higher number of ASB, whereas individuals with higher verbal intelligence were more likely to evade detection, thus being more successful in their antisocial acts. These results sheds further light on the construct of adaptive psychopathy, supporting the notion that also non-incarcerated psychopathic individuals act highly antisocial. Only separate factors like verbal intelligence might mitigate negative consequences. Further implications for the concept of successful psychopathy are discussed.

A field study of acute dynamic risk assessment in individuals convicted of sexual offenses.

Acute dynamic risk factors can change over short periods of time, even within days or hours. When making risk assessments of individuals convicted of sexual offenses, it is necessary to consider this rapidity of change, in order to monitor appropriately the recidivism risk and to prevent future crimes. Until now, there is only one instrument-the Acute-2007-which assesses acute risk factors in individuals convicted of sexual offenses. However, the current state of research about its psychometric properties is still limited. The aim of the present study was to examine the German version of the Acute-2007 in a final sample of N = 534 adult male individuals convicted of a sexual offense in Germany. Within a prospective-longitudinal field study, appropriately trained parole officers used the Acute-2007 directly after a case was allocated to them. By using an average follow-up period of M = 3.83 years, recidivism data were drawn from the Federal Central Criminal Register of the Federal Office of Justice. Results revealed that the Acute-2007 predicted moderately general (AUC = .64), sexual (AUC = .64), and violent (AUC = .64) recidivism. By using Cox regression analysis, the Acute-2007 significantly predicted the time periods until a sexual reoffense. Furthermore, the Acute-2007 added incremental predictive validity beyond the Static-99 in the prediction of sexual recidivism. Taken together, the results of the present study support the relevance of acute dynamic risk factors in risk assessment and management of individuals convicted of sexual offenses. (PsycInfo Database Record (c) 2022 APA, all rights reserved).

Development and preliminary validation of the Emotions while Learning an Instrument Scale (ELIS).

Learning to play a musical instrument is associated with different, partially conflicting emotions. This paper describes the development and psychometric properties of the Emotions while Learning an Instrument Scale (ELIS). In a longitudinal study with 545 German elementary school children factorial structure and psychometric properties were evaluated. Exploratory and confirmatory factor analyses confirmed a two-factor solution measuring Positive musical Emotions while Learning an Instrument (PELI) and Negative Emotions while Learning an Instrument (NELI). Both subscales yielded scores with adequate internal reliability (Cronbach's α = .74, .86) and relatively stable retest reliabilities over 18 months (r = .11 -.56). Preliminary evidence of congruent and divergent validity of the subscales is provided. Implications for future research of musical emotional experiences in children are discussed.

Cross-Validation of the Revised Version of the Violence Risk Appraisal Guide (VRAG-R) in a Sample of Individuals Convicted of Sexual Offenses.

The aim of the present study was to examine the psychometric properties of the German version of the revised Violence Risk Appraisal Guide (VRAG), the VRAG-R. Therefore, VRAG-R ratings were made retrospectively in an Austrian sample of 534 individuals convicted of a sexual offense who were followed up with an average of 7.62 years. The VRAG-R showed large effect sizes for the predictive accuracy of violent (AUC = .75) and general recidivism (AUC = .78) and significant but rather small effect sizes (AUC = .63 and .61, respectively) in predicting any sexual and sexual contact recidivism. Furthermore, for the prediction of violent recidivism but not for sexual recidivism the VRAG-R was incrementally predictive beyond the Sex Offender Risk Appraisal Guide (SORAG) and the Static-99. Finally, the VRAG-R absolute recidivism rates for the risk bins showed satisfactory calibration properties. Taken together, the results of the present study support the cross-national utility of the VRAG-R and its use in applied risk assessment settings also in German-speaking countries.

@myTabu-A Placebo Controlled Randomized Trial of a Guided Web-Based Intervention for Individuals Who Sexually Abused Children and Individuals Who Consumed Child Sexual Exploitation Material: A Clinical Study Protocol.

Background: There is a high demand for evidence-based and cost-effective treatment concepts for convicted individuals who sexually abused children (ISAC) and individuals who consumed child sexual exploitation material (ICCSEM) under community supervision (CS). The @myTabu-consortium developed a guided web-based intervention for convicted ISAC and ICCSEM under CS consisting of six online modules targeting psychological meaningful risk factors. The study aims to evaluate the effectiveness of this guided web-based intervention in reducing dynamic risk factors and the risk to re-offend compared to a placebo condition. Furthermore, these dynamic risk factors are measured before and after every module to evaluate their individual effectiveness to reduce the respective risk factor as well as risk to re-offend. This clinical trial protocol describes the planned methods as well as the intervention concept. Methods: The methodological design is a placebo controlled randomized add-on trial (N = 582) with follow-ups at 8 points in time. The placebo condition controls for attention and expectation effects and comprises the same amount of modules with a comparable temporal effort as the experimental intervention. The trial is conducted as an add-on to community supervision as usually done. Primary outcomes are dynamic risk factors assessed by self-report risk assessment tools and officially recorded re-offenses. Discussion: To the best of our knowledge, the study is the first to compare the (cost-) effectiveness of a guided web-based intervention for convicted ISAC and ICCSEM under community supervision against a placebo condition. Methodological limitations (e.g., potential ceiling- or volunteers-effects) are discussed. Clinical Trial Registration: German Clinical Trial Register (DRKS 00021256). Prospectively registered: 24.04.2020.

Dynamic Risk Assessment of Sexual Offenders: Validity and Dimensional Structure of the Stable-2007.

In this study, the predictive and incremental validity of the Stable-2007 beyond the Static-99 was evaluated in an updated sample of N = 638 adult male sexual offenders followed-up for an average of M = 8.2 years. Data were collected at the Federal Evaluation Center for Violent and Sexual Offenders (FECVSO) in Austria within a prospective-longitudinal research design. Scores and risk categories of the Static-99 (AUC = .721; p < .001) and of the Stable-2007 (AUC = .623, p = .005) were found to be significantly related to sexual recidivism. The Stable-2007 risk categories contributed incrementally to the prediction of sexual recidivism beyond the Static-99. Analyzing the dimensional structure of the Stable-2007 yielded three factors, named Antisociality, Sexual Deviance, and Hypersexuality. Antisociality and Sexual Deviance were significant predictors for sexual recidivism. Sexual Deviance was negatively associated with non-sexual violent recidivism. Comparisons with latent dimensions of other risk assessment instruments are made and implications for applied risk assessment are discussed.

A Validation Study of the German Version of the Sexual Inhibition/Sexual Excitation Scales-Short Form.

The dual control model of human sexual behavior proposes that sexual responses involve an interaction between sexual excitatory and sexual inhibitory systems. The Sexual Inhibition/Sexual Excitation Scales (SIS/SES; Janssen, Vorst, Finn, & Bancroft, 2002) as well as its short form (SIS/SES-SF; Carpenter, Janssen, Graham, Vorst, & Wicherts, 2011) assess individual propensities to become sexually excited and sexually inhibited. The present study utilized a dataset of 2221 German-speaking adults (n = 1214 women and n = 987 men) to examine the psychometric properties of the German version of the SIS/SES-SF. The results of the confirmatory factor analysis indicated that the three factors suggested in the original version can be generally replicated within the current study. Analyses of measurement invariance for gender revealed that the general three-factor structure as well as factor loadings can be regarded as equal for men and women. However, scalar invariance could not be fully obtained, indicating that the intercepts of some items were different for men and women. Taken together, the results of the presents study provide further evidence for the psychometric properties of the German version of the SIS/SES-SF.

The Development and Preliminary Validation of a Brief Questionnaire of Psychopathic Personality Traits.

The measurement of psychopathic personality traits via self-report has become an important tool in legal psychology. One prominent instrument is the Psychopathic Personality Inventory (PPI; Lilienfeld and Andrews, 1996), a well-validated questionnaire that is widely applied in many countries. In Germany, it is the only questionnaire assessing psychopathic traits that is available from a publisher with a manual edited for easy administration. Nevertheless, the PPI shows certain shortcomings: the high number of 154 items makes it less economic, it was developed on a non-representative undergraduate sample, and studies revealed an inconsistent factor structure. To overcome these points, a new questionnaire, the Questionnaire of Psychopathic Personality Traits [German: Fragebogen Psychopathischer Persönlichkeitseigenschaften (FPP)] was developed. The sample consists of n = 132 civilians (56% female) and n = 173 inmates of German correctional facilities (30% female). The FPP comprises 30 items, whose wording was short and adequate for inmates. It shows satisfying psychometric properties regarding factorial structure, item properties, and reliability. Partial invariance regarding both subsamples allows for interpretation of latent means. Results supported validity such as associations with self-reported crime, and inmates' misconduct. The factorial structure was cross-validated on a second sample of N = 517 participants (71% female) from an online study. The FPP is useful in large-scale research studies as well as for clinical settings, e.g., for treatment planning in correctional facilities.