The article examines contemporary approaches to risk assessment in the field of probation within the context of the digital transformation of criminal justice systems. Special attention is paid to information-analytical and algorithmic tools used to support decision-making at the stages of pre-sentence reporting, supervision, and correctional planning. The study provides a comparative analysis of international and national risk assessment models, including the algorithmic system COMPAS (USA), the Ukrainian information subsystem “Kassandra,” and the conceptual Multi-Task Dual-Theory Probation Prediction Model (MT-DT).

The research identifies fundamental differences between actuarial-algorithmic models typical of common law jurisdictions and legally oriented analytical models compatible with continental legal systems. It is substantiated that systems such as COMPAS, which rely on opaque mathematical algorithms and numerical risk indices, raise significant concerns regarding transparency, explainability, procedural fairness, and compliance with the principles of legality and individualization of punishment. These limitations are illustrated through an analysis of the landmark case Loomis v. Wisconsin, which highlighted the risks associated with the use of non-transparent algorithmic assessments in judicial decision-making.

The article argues that the Ukrainian probation system currently relies on information-analytical tools that primarily perform accounting and documentation functions and lack autonomous predictive capabilities. In response to this limitation, the author substantiates the feasibility of implementing an integrated risk assessment model based on the MT-DT concept, which combines formalized legal logic with adaptive predictive algorithms.

Keywords: risk assessment, pretrial report, probation, information-analytical tools, algorithmic tools, committing repeat criminal offenses, recidivism, ECHR practice, forecasting, sentencing, prevention, criminal offense, crime.

References

1. Harcourt, B. E. (2007). Against prediction: Profiling, policing, and punishing in an actuarial age. University of Chicago Press. https://academic.oup.com/chicago-scholarship-online/book/21282
2. O’Neil, C. (2016). Weapons of math destruction: How big data increases inequality and threatens democracy. Crown Publishing Group.
3. Berk, R. A. (2012). Criminal justice forecasts of risk: A machine learning approach. Springer. https://books.google.com.ua/books?id=Jrlb6Or8YisC
4. Berk, R. A. (2019). Fairness in criminal justice risk assessments. Statistical Science, 34(2), 173–183. https://www.cis.upenn.edu/~mkearns/papers/FairnessSMR.pdf
5. Wikipedia contributors. (n.d.). Information and analytical activity. https://uk.wikipedia.org/wiki/Інформаційно-аналітична_діяльність
6. Wang, Q., Zhang, X., Yang, L., Shao, R., Wang, B., Wang, F., & Qu, C. (2025). Incorporating legal logic into deep learning: An intelligent approach to probation prediction. In Proceedings of the 34th International Joint Conference on Artificial Intelligence (IJCAI 2025) (pp. 868–876). https://www.ijcai.org/proceedings/2025/0868.pdf
7. State v. Loomis, 881 N.W.2d 749 (Wis. 2016). Wisconsin Supreme Court. https://law.justia.com/cases/wisconsin/supreme-court/2016/2015ap000157-cr.html
8. Law of Ukraine No. 160-VIII, On Probation (2015, February 5). Bulletin of the Verkhovna Rada of Ukraine, (13), Art. 93.
9. Criminal Code of Ukraine, Law of Ukraine No. 2341-III (2001, April 5). Bulletin of the Verkhovna Rada of Ukraine, (25–26), Art. 131.
10. Criminal Procedure Code of Ukraine, Law of Ukraine No. 4651-VI (2012, April 13). Bulletin of the Verkhovna Rada of Ukraine, (9–13), Art. 88.
11. Criminal-Executive Code of Ukraine, Law of Ukraine No. 1129-IV (2003, July 11). Bulletin of the Verkhovna Rada of Ukraine, (3–4), Art. 21.