Criminalistic potential of machine learning methods in recognizing psychophysiological reactions of participants in investigative interrogations

Abstract

This study provides a comprehensive assessment of the forensic potential of machine learning methods for recognizing and interpreting the psychophysiological reactions of participants in investigative interrogations in the context of the digitalization of criminal proceedings. The research is based on the analysis and synthesis of foreign empirical studies, as well as on comparative and formal-logical methods, and includes a systematic review of contemporary computer vision and deep learning algorithms used to analyze facial microexpressions and nonverbal behavior. Empirical evidence from studies conducted by Chinese, American, and
Dutch researchers demonstrates that intelligent algorithms are capable of reliably detecting facial microdynamics, indicators of anxiety, and emotional instability, in some cases exceeding the consistency of human observation. At the same time, the findings indicate that algorithmic outputs cannot function as autonomous sources of evidence. The study substantiates the need for normative and methodological adaptation of these technologies to domestic forensic practice in the Republic of Kazakhstan, taking into account evidentiary standards, procedural safeguards, and human rights protection. It is concluded that the phased and explainable integration of machine-based psychophysiological analysis into existing forensic psychological and criminological examinations is both feasible and methodologically justified.