Structural bias and geographical imbalances in AI ethics: A review

Abstract

This study conducts a systematic literature review (SLR) and bibliometric analysis on global research in artificial intelligence (AI) ethics. The study systematically synthesizes the latest developments and focuses on the thematic area in the field by drawing on the PICOC protocol and the PRISMA framework. Above all, it underscores that algorithmic bias arises from embedded social structures, institutional power relations, and data limitations. Additionally, it demonstrates that academic production is geographically imbalanced, with the North overwhelmingly producing the academic output, and the South contributing at a very low level. To focus on key areas for improvement, this study utilizes literature screening and data visualization techniques to identify research gaps and recommends a paradigm shift from equity adjustment at the algorithmic level to a broader sociotechnical system reconstruction.