ALGORITMI VEŠTAČKE INTELIGENCIJE ZA „OCENU RIZIKA“ U DOMENU SAVREMENOG PROGNOZIRAJUĆEG RADA POLICIJE
ARTIFICIAL INTELLIGENCE ALGORITHMS FOR “RISK ASSESSMENT” IN THE FIELD OF MODERN PREDICTIVE POLICING

Summary/Abstract: The work of “risk assessment” algorithms within the framework of predictive policing is based on a risk-based, future-oriented approach, risk generation, and a specific type of police intervention towards individuals. Modern algorithms use AI elements to enhance police efficiency and support more informed decision-making, but they also carry numerous inherent risks to the protection of fundamental human rights and freedoms. The significance of this topic stems from the rapid development of AI systems, the practices of applying AI algorithms in Europe and global, and the recent formulation of international and European legal frameworks for the use of AI systems, which also impact the field of predictive policing. The author starts from the fundamental hypothesis that the international and European legal framework does not introduce an apsolute ban on the use of AI systems in predictive policing but essentially establishes a limited prohibition and introduce basic safeguards for the application of AI systems, as a result of balancing the expansion of AI technology development, its benefits and risks. The aim of the research is to highlight the need for a certain degree of redefinition of the fundamental institutes and principles of criminal procedural law, guarantees of the right to liberty and security, and the requirements of the right to a fair trial, as well as to identify the principles and basic guidelines for the application of AI systems in predictive policing. Two important principles for the application of AI algorithms are: 1. the essential/significant/dominant role of humans in decision-making, which implies that a human carefully evaluates the algorithm’s report by assigning its limited value in decision-making or not assigning it any value at all (depending on the algorithm’s shortcomings or the absence of safeguards for the use of AI systems); 2. the transparency of the AI algorithm’s operation concerning an individual’s right to review the data, functioning, and outcome of the algorithm. The main research results are reflected in linking the functioning of predictive algorithms with criminal procedural law, the principle of adversarial proceedings/contradictory, the right to liberty and security, the right to fair procedure, and identifying fundamental safeguards and risks highlighted by comparative legal practice and international legal documents.

• Details
• Contents