A Framework for Defining Algorithmic Fairness in the Context of Information Access

ABSTRACT As technologies powered by Artificial Intelligence (AI) and Machine Learning (ML) algorithms increasingly take over personal computing online and public sector domains, they simultaneously raise the promise of an extensively productive and sustainable future, as well as fears of widening inequalities, information and content divide, and a more complex information‐seeking landscape. Thus, the hopes of improved accuracy, efficiency, productivity, reduced human bias in decision‐making, and access to information are fast giving way to a trove of ethical and human rights issues with far‐reaching consequences for accountability, privacy, social justice, equity, inclusion, and informed consent, and public participation in decision‐making. Since no technology is entirely free of bias, this paper identifies algorithmic fairness as a more realistic threshold and goal. Building on findings from a previous PRISMA review of relevant literature, the paper proposes a comprehensive framework for defining algorithmic fairness in the context of information access.