Predictive Power of Dynamic (vs. Static) Risk Factors in the Finnish Risk and Needs Assessment Form

Abstract

We estimated the predictive power of the dynamic items in the Finnish Risk and Needs Assessment Form ( Riski- ja tarvearvio [RITA]), assessed by caseworkers, for predicting recidivism. These 52 items were compared to static predictors including crime(s) committed, prison history, and age. We used two machine learning methods (elastic net and random forest) for this purpose and compared them with logistic regression. Participants were 746 men who had and 746 who had not reoffended during matched follow-up periods from 0.5 to 5.8 years. Both RITA items and static predictors predicted general and violent recidivism well (area under the curve [AUC] = .74-.78), but to combine them increased discrimination only slightly over static predictors alone (ΔAUC = .01-.03). Calibration was good for all models. We argue that the results show strong potential for the RITA items, but that development is best focused on improving usability for identifying treatment targets and for updating risk assessments.

Machine Learning and Criminal Justice: A Systematic Review of Advanced Methodology for Recidivism Risk Prediction

“…Both cohorts were tested with ML models, but for this review we consider only the 1977–1978 data excluding the Tollenaar’s 2019 study [ 32 ] because the 1980 cohort showed a worse calibration probability. The dynamic elements of the Finnish Risk and Needs Assessment Form (RITA) were used by Salo and colleagues [ 27 ] to predict general and violent recidivism. The sample included 746 men sentenced to a new term of imprisonment.…”

“…The sorting leads to four categories: general [ 22 , 23 , 24 , 28 , 31 , 32 ], sexual [ 31 , 32 ], violent [ 22 , 25 , 31 ] and all other recidivism. The last category includes studies that considered a specific type of crime [ 21 ] or referred only to males [ 27 ] or youth [ 26 , 29 , 30 ].…”

“…The most used ML model is the logistic regression [ 21 , 30 , 31 , 32 ] and two of its variants, the LogitBoost [ 22 ] and the generalized linear models with ridge and lasso regularization (Glmnet) [ 24 ]. The second most popular model is the random forest [ 23 , 26 , 27 , 33 ]. The other ML models to mention are multi-layer perceptron (MLP) [ 25 ], linear discriminant analysis (LDA) [ 31 ] and penalized LDA [ 32 ].…”

“…The Thailand dataset has a particular relevance with an ACC of 0.90 obtained with logistic regression [ 21 ]. In term of AUC, a model obtained a significant score of 0.78 using the RITA+ dataset trained with random forest [ 27 ].…”

Socioeconomic Effects in Cyberbullying: Global Research Trends in the Educational Context

“…There are also previous researches of CB behavior based on general criminological theories. Three of these theories have a greater theoretical incidence on personal factors: the Theory of Tension, by Robert Agnew; the Social Cognitive Theory with its mechanisms of moral disassociation, by Albert Bandura; and Ajzen's Theory of Reasoned Action; and a fourth that affects aspects of opportunity: the Theory of Daily Activities [43][44][45].…”