Enabling Fairness in Healthcare Through Machine Learning

Abstract

The use of machine learning systems for decision-support in healthcare may exacerbate health inequalities. However, recent work suggests that algorithms trained on sufficiently diverse datasets could in principle combat health inequalities. One concern about these algorithms is that their performance for patients in traditionally disadvantaged groups exceeds their performance for patients in traditionally advantaged groups. This renders the algorithmic decisions unfair relative to the standard fairness metrics in machine learning. In this paper, we defend the permissible use of affirmative algorithms; that is, algorithms trained on diverse datasets that perform better for traditionally disadvantaged groups. Whilst such algorithmic decisions may be unfair, the fairness of algorithmic decisions is not the appropriate locus of moral evaluation. What matters is the fairness of final decisions, such as diagnoses, resulting from collaboration between clinicians and algorithms. We argue that affirmative algorithms can permissibly be deployed provided the resultant final decisions are fair.

Health equity assessment of machine learning performance (HEAL): a framework and dermatology AI model case study

“…Other work has emphasised the need for operationalizing equity and fairness in AI for healthcare. 11 , 12 Recent work addressing this call has discussed ethical considerations of fairness and equity in the context of AI for healthcare, 42 , 43 , 44 , 45 , 46 , 47 suggested best practices to incorporate health equity in the algorithm development lifecycle, 14 , 26 , 27 , 28 , 29 , 48 and proposed operational definitions. 30 , 31 , 32 Existing operational definitions have largely borrowed from the AI fairness literature, 21 , 22 , 23 , 24 proposing metrics based on statistical parity in AI performance across subpopulations.…”

Weight Perturbation Can Help Fairness under Distribution Shift

“…Fairness in Machine Learning. Fairness [Mehrabi et al, 2021, Pessach and Shmueli, 2022, Madaio et al, 2022, Li and Varakantham, 2022, Padh et al, 2021, Jiang et al, 2022a] is a legal requirement for machine learning models for various high-stake real-world predictions, such as healthcare [Ahmad et al, 2020, Bjarnadóttir and Anderson, 2020, Grote and Keeling, 2022, education [Bøyum, 2014, Brunori et al, 2012, Kizilcec and Lee, 2022, and job market [Hu and Chen, 2018, Alder and Gilbert, 2006]. Achieving fairness in machine learning is a challenging problem, such as bias, and discrimination mitigation in datasets.…”

Ethical assessments and mitigation strategies for biases in AI-systems used during the COVID-19 pandemic

“…Algorithmic injustice arises when the results of automated algorithmic decision-making powered by machine learning lead to unjustified, unfair, and discriminatory outcomes (Grote and Keeling, 2022;Hedden, 2021;Marjanovic et al, 2021). Impact Machine-learning algorithms have numerous impacts in terms of which cases are considered alike and how others are treated.…”