Mitigating Bias in Machine Learning Models with Ethics-Based Initiatives: The Case of Sepsis.

Abstract

This paper discusses ethics-based strategies for mitigating bias in machine learning models used to predict sepsis onset. The first part discusses how various kinds of bias and their potential synergies can reduce predictive accuracy, especially as those biases derive from social determinants of health (SDOHs) and from the design and construction of the predictive model. The second part of the essay discusses how certain ethically-based strategies might mitigate the potential for disparate or unfair treatment produced by these models, not only as they might apply to sepsis but to any syndrome that witnesses the impact of adverse SDOHs on socioeconomically disadvantaged or marginalized populations.