Artificial intelligence as a social phenomenon: A bibliometric reading through the lenses of structural violence, intersectionality, and surveillance

Abstract

This study investigates how social vulnerabilities associated with the use of artificial intelligence (AI) in medicine are reflected in recent biomedical literature and how these patterns correlate with central theoretical directions in sociology and social work. Through a bibliometric analysis of 2,589 meta-analyses and systematic reviews published between 2020 and 2025 in PubMed, the research maps the conceptual structure of the field using co-occurrence networks at two thresholds (5 and 20). The results show a concentration of discussions on the technical and clinical aspects of AI (diagnosis, predictive modelling, electronic health records, large language models), while terms expressing social and ethical concerns (equity, algorithmic bias, privacy, ethics, health disparities, clinical competence) occupy semi-peripheral positions in the network. Interpreting these structures through theoretical lenses such as structural violence, social determinants of health, intersectionality, algorithmic oppression, surveillance capitalism, and care ethics reveals that AI risks reproducing and intensifying pre-existing inequalities. The analysis emphasises that algorithmic bias, unequal data infrastructures, model opacity, and changes in the distribution of clinical work are not isolated phenomena, but manifestations of broader social processes that shape vulnerability and exclusion. Therefore, the study argues for the need to integrate sociological and social work perspectives into the development and evaluation of medical AI and advocates for interdisciplinary approaches that place equity, transparency, and the experiences of marginalised populations at the centre. Such an orientation is essential for AI in medicine to contribute to reducing — rather than amplifying — social inequalities in health.