A drawing is worth a thousand words : analyzing children’s projective techniques with machine learning
Date
2024-12
Authors
Banfi, Catalina
relationships.isContributorOfPublication
Edo, María
Sosa Escudero, Walter
Journal Title
Journal ISSN
Volume Title
Publisher
Universidad de San Andrés. Departamento de Economía
Abstract
Child maltreatment is a critical global issue with long-lasting physical, psychological, and social consequences. A major challenge in addressing this problem is substantial underreporting, as most cases occur within the home, preventing timely intervention and leaving many children without support. While detection tools exist, they are typically applied only to cases already within the system, resulting in a reactive rather than proactive approach to identifying at-risk children. The main objective of this thesis is to develop a machine learning algorithm to proactively enhance the detection of child maltreatment through the analysis of projective techniques applied to primary school children. Using data from an Inter-American Development Bank project, the study automates the evaluation of graphical indicators in children’s drawings, simulating psychologists’ cognitive processes to generate risk alerts and identify cases requiring further psychodiagnostic evaluation. The resulting algorithm achieved high accuracy metrics, enabling large-scale implementation while maintaining efficiency. The intervention costs approximately $2 USD per child and significantly reduces future costs associated with the long-term consequences of child maltreatment. This proposal, designed as a public policy for primary schools, trains teachers to identify key indicators in drawings, embedding detection into the educational system and optimizing resource allocation.
Description
Fil: Banfi, Catalina. Universidad de San Andrés. Departamento de Economía; Argentina.