Abstract

This study presents an integrated socio-technical and data-driven framework to model the psychological and contextual determinants of donor behavior in disaster settings. From an artificial intelligence perspective, the research introduces a hybrid architecture that combines multigroup, multilevel, and confirmatory structural equation modeling (SEM) with a semi-supervised autoencoder-based clustering strategy for latent profile identification. A MIMIC (Multiple Indicators Multiple Causes) extension further links latent constructs to observable donation frequency, bridging attitudinal and behavioral dimensions of prosocial action. From an engineering and humanitarian logistics perspective, this framework enhances the prediction and segmentation of donor behavior under disaster exposure. The model, based on survey data from disaster-affected individuals in Chile, incorporates validated constructs such as social vulnerability, solastalgia (emotional distress from environmental degradation), deprivation cost, resilience, and climate-related anxiety. Five distinct psychological donor profiles were identified, behaviorally validated, and linked to operational implications for mitigating material convergence and optimizing resource allocation. The findings offer a socio-technical pathway to integrate psychosocial assessment with humanitarian logistics decision-making, advancing anticipatory capacity, behavioral forecasting, and equitable supply distribution. The study contributes a replicable framework that connects human-centered behavioral modeling with operational optimization, supporting more adaptive and resilient disaster response systems