Artificial intelligence-driven approaches for assessing social vulnerability to natural hazards: a comprehensive bibliometric review

Official URL: https://dx.doi.org/10.1007/s11069-025-07808-w

Artificial intelligence (AI) applications in social vulnerability assessments have gained increasing attention in disaster management and risk reduction research. This study presents a comprehensive bibliometric analysis of AI applications in social vulnerability assessments from 1996 to 2024, analysing 26,717 publications from Scopus and Web of Science databases. The field has grown rapidly, with an annual publication growth rate of 19.94% and an average of 33.05 citations per document, reflecting both scale and scholarly impact. The analysis further highlights the collaborative nature of the field, with 63,778 contributing authors, an average of 4.94 co-authors per publication, and 27.69% of outputs featuring international co-authorship. Publication types are dominated by journal articles (25,910), supplemented by reviews (612) and conference papers (195). Leading publication venues include Remote Sensing and IEEE Transactions on Geoscience and Remote Sensing, while China, the United States, and India emerge as the most prolific contributors, alongside growing participation from developing economies. Keyword analysis across three distinct periods, Foundational (1996–2009), Developmental (2010–2019), and Transformative (2020–2024), shows a shift from traditional geospatial methodologies to advanced AI-driven analytics, with terms such as “machine learning,” “deep learning,” and “synthetic data” emerging as dominant themes. Remote sensing, GIS, and decision support systems have remained central, while recent years have focused more on convolutional neural networks and semantic segmentation. Overall, our review provides one of the largest, periodized mappings of this domain and surfaces critical gaps (e.g., under-integration of social science) while offering an actionable agenda centred on explainable AI, multimodal data fusion, participatory approaches, and responsible/ethical AI to enhance decision-making for disaster resilience.