Artificial Intelligence in Criminalistics and Forensic Examination: Issues of Legal Personality and Algorithmic Bias

Abstract

Active development and implementation of artificial intelligence technologies (AI) in various spheres of human activity have started the processes of qualitative change in public relations. This fact necessitates the development of legal and technical standards to regulate AI technologies. In this regard, the most controversial issue is the recognition of AI personality. The analysis of various opinions on the matter shows the lack of a consolidated approach in the existing legal doctrine. Creating the legal status for AI systems would provide for several options depending on its type and purpose – from technical means to the status of an “electronic personality” and recognition as a full-fledged subject of law.
Considering the specifics of criminalistics and forensic examination, it is better to position AI systems as technical means. Machine learning is considered a form of AI. It is the use of mathematical data models that enables computer training through specialized algorithms and training data. Algorithms can create or reproduce distortions and inaccuracies unintentionally embedded in the training data, which causes the manifestation of algorithmic bias. To eliminate bias of algorithms it is necessary to pay attention to the quality of training data. The author has developed special methods to prepare such data, which are presented in this article in relation to ballistic identification systems. Also, one of the elements of system technical solutions to the problem of bias of AI algorithms is the development of standards for minimizing unjustified bias in algorithmic solutions.