Algorithmic fairness: a complexity science perspective for public health research and practice

Algorithmic fairness research has largely emphasized model-level metrics, despite evidence that algorithmic systems behave differently once deployed in real-world, complex health systems. This paper reframes fairness as a system-level property and argues that integrating complexity science, implementation research, and sociotechnical perspectives is essential to addressing a central unresolved question: whether fairness-informed algorithms can sustainably narrow health inequalities over time.