HMCI-Net: Explainable Artificial Intelligence for Cognitive Health and Historical Trauma Awareness Analytics in Partition Cinema

Abstract

Historical memory generated through cinema is commonly assessed using descriptive survey analysis, which provides limited support for computational modeling and educational decision-making, even though exposure to collective-trauma narratives carries measurable cognitive-health and psychosocial consequences for audiences. This paper proposes HMCI-Net, an Explainable Artificial Intelligence (XAI) framework for computational modeling of historical memory cognition through Partition cinema. The framework introduces a novel Reliability–Entropy Adaptive Fusion (REAF) algorithm that estimates adaptive construct weights by jointly considering psychometric reliability, information diversity, and construct uniqueness. These weights are integrated to compute the Historical Memory Cognition Index (HMCI), a multidimensional measure representing historical awareness, emotional empathy, perceived authenticity, national integration, and educational impact. These constructs are treated as cognitive-health and trauma-awareness indicators rather than clinical diagnostic measures. The framework further incorporates explainable machine learning and scenario-based simulation to support educational decision-making. Experimental evaluation was conducted on 160 valid audience responses using 1,000 bootstrap iterations and 5,000 Monte Carlo simulations. The proposed REAF algorithm selected exponent values of (0.5, 0.5, 0.5) and produced balanced adaptive weights ranging from 0.1703 to 0.2308. The resulting HMCI achieved a mean of 0.6487, while the dataset demonstrated strong psychometric adequacy (KMO = 0.915; Bartlett's χ² = 3343.79, p < 0.001). Comparative analysis showed that REAF generated more interpretable and statistically robust construct weights than conventional equal-weight, entropy-based, and PCA-based methods. The simulation module further quantified the expected improvement in historical memory under alternative educational interventions. The proposed framework offers a transparent, reproducible, and explainable computational approach for cognitive-health oriented audience analytics, health humanities, and trauma-informed educational decision support.