FEATURES OF APPLICATION OF MACHINE LEARNING TOOLS FOR PSYCHOLOGICAL ASSESSMENT OF CHILDREN'S DRAWINGS

The article presents an analysis of the features of the application of projective methods for testing the psychological state of children, as well as a pilot experiment on the implementation of machine learning tools for the psychological assessment of their drawings. These issues must be considered comprehensively, since most methods of psychological assessment and, in particular, projective technologies impose special requirements for reliability and validity, and also require special training and professional competence for their interpretation. The goal of the research project is to use digitized drawings created by internally displaced children of Nagorno-Karabakh as a tool to identify and assess their psychological state using machine learning methods. We believe that this approach can provide valuable insights into the mental well-being of children affected by the conflict and can serve as a diagnostic tool in times of economic and humanitarian crisis. The proposed feature extraction approach in psychological assessments identifies key visual elements such as object size, color, and spatial relationships that provide insight into emotional and cognitive states. Therefore, the application of machine learning techniques, and in particular feature extraction, classification, clustering, and the use of large language models (LLMs), offers a transformative approach to the psychological assessment of children's drawings. Using neural networks (NNs) for feature extraction, these models can automatically detect important visual elements such as line intensity, shape proportions, color use, and symbolic objects. This automated extraction reduces the subjectivity and variability associated with traditional manual interpretation, improving the accuracy and scalability of assessments. Furthermore, classification and clustering algorithms provide structured ways to group children's drawings based on similar psychological themes, allowing for more objective analysis.