Artificial Intelligence for Poverty Analytics: A Bibliometric and Systematic Review of Prediction, Mapping, and Governance Models

As proclaimed by the UN, artificial intelligence (AI) is seen as a possibly useful strategic tool in poverty reduction worldwide under the aegis of Sustainable Development Goals (SDGs). However, current research is scattered and involves predictive modelling, geospatial analysis and angles of governance, making it impossible to achieve a comprehensive view on AI's effective operation in different social and institutional contexts. This study, using bibliometric analysis combined with the systematic review, is based on PRISMA methodology of all articles indexed in Scopus over AI-related research to map intellectual and methodological structures. In all, the results illustrate that poverty research based on AI will mainly concern three main areas: machine learning of poverty prediction, GIS mapping with AI, and the relationship between AI, poverty, and governance. This paper aims to analyse these intersections by linking the three areas together. While making some gains in this area, we still find that much of the literature exists in disciplinary "silos", policy analysis is disconnected from its implementation details and there's no consideration of qualitative factors such as what sorts should come first and whether or not the information infrastructure was already heralded by changes elsewhere in society. To resolve such problems, the study argues for an integrative framework that takes AI as a sociotechnical system within governance ecosystems. This stance provides new insight on how AI can best support targeted interventions, resource optimization, and adaptive policies. The paper is a contribution to the consolidation of research in fragmented areas and to promotion of a strong theory. More specifically, it supports the idea that data-based methods should be linked with institutional abilities in order to produce effective poverty reduction outcomes.