Insights From Insurance for Fair Machine Learning
Christian Fröhlich, Robert C. Williamson
TL;DR
The paper argues that insurance offers a fruitful analogy for understanding the social situatedness of fair machine learning, moving beyond purely statistical notions of fairness. It develops a historical and normative analysis of insurance rationalities—solidarity, actuarial fairness, and personalization—and links these to ML concepts like calibration and independence. A central contribution is the fourfold responsibility framework (causal, control-based, moral, material) and the emphasis on normative questions about what is within an individual's control, informed by real-world practices like smoking, genetics, and behavioral data. Additionally, the work foregrounds performativity and the aggregate-versus-individual tension, arguing for reflexive, context-aware design and evaluation of ML systems that account for social impact and potential feedback effects.
Abstract
We argue that insurance can act as an analogon for the social situatedness of machine learning systems, hence allowing machine learning scholars to take insights from the rich and interdisciplinary insurance literature. Tracing the interaction of uncertainty, fairness and responsibility in insurance provides a fresh perspective on fairness in machine learning. We link insurance fairness conceptions to their machine learning relatives, and use this bridge to problematize fairness as calibration. In this process, we bring to the forefront two themes that have been largely overlooked in the machine learning literature: responsibility and aggregate-individual tensions.