Conceptualizing the Ethical Implications of Bias in Artificial Intelligence Systems: A Systematic Literature Review

Abstract

Artificial intelligence (AI) systems are increasingly mediating consequential decisions in healthcare, finance, employment, and criminal justice, yet a growing body of evidence shows that these systems can encode and amplify social bias, leading to unfair, discriminatory, and ethically problematic outcomes. The literature in this area has increased rapidly, but it is fragmented across computer science, ethics, law, and the social sciences, and few studies synthesize it using a clear and reproducible methodology. This research tackles this gap using a systematic literature review (SLR) of 110 peer-reviewed publications published between 2016 and 2023, selected and reported following PRISMA 2020 principles. A search was performed in Scopus for studies that were filtered according to predetermined inclusion and exclusion criteria and analyzed by thematic content analysis. The synthesis identifies three major sources of bias (data-driven, algorithmic and human/cognitive) and four common ethical features (fairness and equity, transparency and accountability, governance and regulation and mitigation). The research finds that datadriven bias and fairness concerns dominate the literature, mitigation methods are largely technical and seldom verified in implementation, and governance measures trail behind technical advancements. The paper contributes a coherent conceptual framework linking sources of prejudice with ethical implications and levels of mitigation, and a description of an agenda for future empirical research.