Università degli Studi eCampus

The work presented herein proposes DisasterReliefGPT, a multimodal AI system for automation in the areas of crisis communication and post-disaster assessment. The system integrates three tightly coupled components: a vision module called DisasterOCS for structural damage detection in satellite images, a Large Vision–Language Model (LVLM) for enhanced visual understanding and contextual reasoning, and a Large Language Model (LLM) to produce detailed, clear assessment reports. DisasterOCS relies on a ResNet34-based encoder with partial weight sharing and event-specific decoders, coupled with a custom MultiCrossEntropyDiceLoss function for multi-class segmentation on pre- and post-disaster image pairs. On the benchmark xBD dataset, the developed system reaches a high score of 78.8% in identifying F1-damage, making correct identifications of destroyed buildings with 81.3% precision, while undamaged structures are found with a very high value of 90.7%. From a combination of these components, emergency responders can immediately provide reliable and readable assessments of damage that can be used to directly support urgent decision-making.

Abstract Employment inclusion for people with disabilities remains a persistent challenge across developed nations, with employment rates 24 percentage points below general population levels in Europe. Traditional centralized employment matching systems face privacy constraints, data sovereignty requirements, and limited collaborative learning capabilities. This paper presents a comprehensive privacy-preserving federated learning framework for multi-regional disability employment matching that enables collaborative model training while maintaining data locality and regulatory compliance. Our approach combines ensemble-based LightGBM federation with parameter-level MLP federation, achieving 0.9011 F1-score (LightGBM) and 0.7881 F1-score (MLP with privacy preservation) across five Italian employment centers in Veneto region. The privacy framework integrates differential privacy with RDP composition (ε = 1.0, δ = 10⁻⁶), Shamir’s 3-of-5 secret sharing for secure aggregation, and blockchain anchoring for long-term integrity verification. Empirical evaluation demonstrates minimal performance degradation under federated constraints (0.0005 F1-score loss) while providing formal privacy guarantees. The system reduces manual processing time from 30–60 min to under 5 min per candidate with sub-100 ms response times suitable for real-world deployment. Ongoing pilot deployment at partner employment centers validates practical applicability for European disability employment services. The research contributes to trustworthy AI frameworks for sensitive social applications requiring regulatory compliance, algorithmic fairness, and privacy preservation.

This study explored the influence of different counter movement jump (CMJ) goals on performance, kinetics, and kinematics between 56 highly-trained and elite track and field (T&F), football, and futsal athletes. Within and between-sport difference were compared when aiming to: (a) "jump as high as possible" (CMJh) and (b) "jump as fast as possible" (CMJf), using a mixed MANOVA and follow-up univariate mixed ANOVAs. Movement goals had a significant main effect on all variables (p < 0.001). Comparatively, CMJf elicited higher mean propulsive power normalized to body mass (MPPbm, ηp2 = 0.794) and reactive strength index modified (RSIMOD, ηp2 = 0.755), alongside lower jump height (ηp2 = 0.782), contraction time (ConT, ηp2 = 0.857), propulsive displacement (ηp2 = 0.751), and countermovement velocity (CMvelocity, ηp2 = 0.600). Sport interaction analyses revealed that T&F athletes consistently outperformed the other sports in RSIMOD across both movement goals (d = 1.105 - 1.598). MPPbm and ConT differed significantly between T&F and football (d = 0.947 - 1.324). A Movement goal × Sport interaction for CMvelocity indicated that T&F and football athletes increased CMvelocity under CMJh (d = 1.242 - 1.635, p < 0.001) compared to CMJf, whereas futsal players maintained similar downward motion across goals (p = 0.938). In conclusion, movement goals significantly modify CMJ performance variables in such athletes, and these effects are further influenced by sport specialization. Clear and goal-specific verbal instructions should be standardized in CMJ testing to ensure reliable athlete monitoring and performance evaluation.

Gut dysbiosis, an imbalance in gut microbiota, is increasingly linked to depression through the microbiota-gut-brain axis. Stress is an important risk factor for both gut dysbiosis and depression. Despite evidence of altered gut microbiota composition in patients with depression, little is known about how stress and specific gut microbiota features interact to influence depressive symptoms in healthy individuals. This study examined 398 healthy adults (241 women) who provided stool samples and completed validated questionnaires on perceived stress (PSS) and depressive symptoms (CES-D), with a focus on sex differences. In this sample, men and women were characterised by similar gut microbiota composition and diversity. Women reported higher PSS scores than men, whereas no differences were found in CES-D scores. Using Bayesian analyses, results showed that perceived stress predicted depressive symptoms in both sexes. Notably, in women, the genus Eubacterium moderated this relationship: higher perceived stress combined with lower Eubacterium abundance predicted more severe depressive symptoms. In contrast, no moderations by gut bacteria were found in men. The current results warrant further sex-specific investigations of the interaction between stress and specific gut microbiota features in influencing depressive symptoms and suggest that the genus Eubacterium might be a promising microbial biomarker associated with depressive symptoms, particularly in women under higher stress levels.

The Existential Quest Scale (EQS) is a brief instrument designed to assess individuals’ willingness to engage with existential quest in both religious and secular contexts. As the construct of existential quest becomes increasingly relevant for understanding psychological flexibility, identity development, and social attitudes in multicultural societies, ensuring the validity of its measurement is essential. Previous validations of the EQS have relied on Classical Test Theory (CTT), which limits comparability across groups and item-level precision. This study aims to evaluate the EQS using Rasch modeling, a robust item response theory (IRT) approach that overcomes such limitations. Drawing on a large, heterogeneous sample (N = 4,378), we assessed dimensionality, item functioning, and measurement invariance across sex, age, and religious affiliation. Results confirmed the unidimensional structure of the EQS, its ability to discriminate levels of existential quest, and its psychometric invariance across demographic groups. Findings also suggested revisions to item and response category functioning to enhance scale performance. By applying IRT to the EQS, this study advances the psychometric assessment of complex, culturally adaptive constructs and supports the EQS as a rigorous tool for research and applied settings.