Abstract
This article examines the application of algorithmic recidivism risk assessment systems in parole procedures within a comparative framework. It analyzes the experience of Council of Europe member states (the Netherlands, Germany, Sweden, and the United Kingdom) alongside the current legal regulation of parole in the Republic of Kazakhstan. The aim of the study is to substantiate the architecture of a national risk assessment algorithm integrated with the information systems “Torelik”, the Committee on Legal Statistics and Special Records of the Prosecutor General’s Office (KSPSU), and the Automated Information System of Enforcement Proceedings (AIS OIP) in order to verify the fulfillment of civil obligations. The methodological framework of the research is based on comparative legal, statistical, and formal legal methods. The analysis of European legislation and judicial statistics of the Supreme Court of the Republic of Kazakhstan for 2020–2025 demonstrates that the Kazakhstani parole model relies on a system of behavioral grading of convicted persons, which has become a mandatory element in the preparation of institutional reports submitted to the court when considering parole applications. However, these indicators are characterized by low predictive accuracy and a high degree of subjectivity in their formation. Based on the findings, recommendations are formulated for the phased adaptation of algorithmic risk assessment tools to the conditions of the Kazakhstani legal system. It has been established that, unlike its European counterparts, the Kazakhstani model lacks both a validated tool for predicting recidivism and an appellate mechanism for reviewing such assessments, resulting in substantial regional variation in parole approval rates. The practical significance of the study lies in proposing a phased roadmap for integrating algorithmic risk-assessment tools into Kazakhstan’s legal system while preserving the primacy of judicial discretion.