Machine Learning Models for Behavioral Analysis in Criminal Investigation Training
Abstract
Abstract— One of the vital aspects of criminal investigation training is behavioral analysis that influences the capacity of criminal investigators to perceive verbal and non-verbal indications, interview management and decision-making in the context of uncertainty. The old fashioned approaches to training are extensively relying on the subjective evaluation of the instructor which in most cases is inconsistent and unmeasurable. The given paper suggests an idea of a machine learning-based system of behavior analysis within the criminal investigation training, founded on multiproteic data including interview transcripts, characteristics of a speech, and communication logs. Sequential learning models are used to represent the temporal nature of the investigative action process and explainability aligned to rubrics is used to provide transparent and actionable feedback to trainees. Experimental results show that experimental multimodal approach can significantly enhance the accuracy of assessment, training and bias avoidance compared with the conventional and single modality methods. The results call for the use of machine learning in the context of supporting objective, ethical and adaptive training environments in the process of building future-ready criminal investigation training systems.