BSBenny Salo

Benny Salo

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Benny Salo has a total of 11 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

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

    by Benny Salo, Toni Laaksonen, Pekka Santtila

    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.


  • Construct validity and internal reliability of the Finnish risk and needs assessment form

    by Benny Salo, Toni Laaksonen, Pekka Santtila

    Validation of risk and needs assessment instruments used to predict, and reduce, recidivism and misconduct is of ethical, practical and scientific importance. We argue for a focus on variable (i.e. changeable over time) risk factors, and that validation begins with establishing construct validity. The Finnish Risk and Needs Assessment Form, in Finnish Riski- ja tarvearvio (RITA), is a semi-structured interview form adapted from the Offender Assessment System used in England and Wales and consists primarily of variable risk factors. In this study, we examined the construct validity and internal reliability of RITA. The results suggest that the original RITA sections do not provide an adequate statistical model for describing the relationship between scored questions, and we, therefore, offer an improved model. In our proposed model, several scores are influenced by more than one dimension of risk and the dimensions correlate with each other considerably. We suggest that the dimensions that can be measured with RITA are Problems managing economy, Alcohol problems, Resistance to change, Drug abuse and associated behaviour, Aggressiveness and Employment problems. All factors except Drug abuse and associated behaviour had very good internal reliability. We propose this new model to be used in future research with, and in development of, RITA.


  • Features used by judges to evaluate expert witnesses for psychological and psychiatric legal issues

    by Alessandro Tadei, Katarina Finnilä, Julia Korkman, Benny Salo, Pekka Santtila


  • Using Bayes’ theorem in behavioural crime linking of serial homicide

    by Benny Salo, Jukka Sirén, Jukka Corander, Angelo Zappalà, Dario Bosco, Andreas Mokros, Pekka Santtila

    Purpose. The study extends research by Santtila et al. (2008) by investigating the effectiveness of linking cases of serial homicide using behavioural patterns of offenders, analysed through Bayesian reasoning. The study also investigates the informative value of individual behavioural variables in the linking process. Methods. Offender behaviour was coded from official documents relating to 116 solved homicide cases belonging to 19 separate series. The basis of the linkage analyses was 92 behaviours coded as present or absent in the case based on investigator observations on the crime scene. We developed a Bayesian method for linking crime cases and judged its accuracy using cross‐validation. We explored the information added by individual behavioural variables, first, by testing if the variable represented purely noise with respect to classification, and second, by excluding variables from the original model, one by one, by choosing the behaviour that had the smallest effect on classification accuracy. Results. The model achieved a classification accuracy of 83.6% whereas chance expectancy was 5.3%. In simulated scenarios of only one and two known cases in a series, the accuracy was 59.0 and 69.2%, respectively. No behavioural variable represented pure noise but the same level of accuracy was achieved by analysing a set of 15, as analysing all 92 variables. Conclusion. The study illustrates the utility of analysing individual behavioural variables through Bayesian reasoning for crime linking. Feasible applied use of the approach is illustrated by the effectiveness of analysing a small set of carefully chosen variables.