A Review on Fairness in Machine Learning

Abstract

An increasing number of decisions regarding the daily lives of human beings are being controlled by artificial intelligence and machine learning (ML) algorithms in spheres ranging from healthcare, transportation, and education to college admissions, recruitment, provision of loans, and many more realms. Since they now touch on many aspects of our lives, it is crucial to develop ML algorithms that are not only accurate but also objective and fair. Recent studies have shown that algorithmic decision making may be inherently prone to unfairness, even when there is no intention for it. This article presents an overview of the main concepts of identifying, measuring, and improving algorithmic fairness when using ML algorithms, focusing primarily on classification tasks. The article begins by discussing the causes of algorithmic bias and unfairness and the common definitions and measures for fairness. Fairness-enhancing mechanisms are then reviewed and divided into pre-process, in-process, and post-process mechanisms. A comprehensive comparison of the mechanisms is then conducted, toward a better understanding of which mechanisms should be used in different scenarios. The article ends by reviewing several emerging research sub-fields of algorithmic fairness, beyond classification.

Fairness in Mobile Phone–Based Mental Health Assessment Algorithms: Exploratory Study

“…There, a trade-off is expected between accuracy and fairness (ie, with increased fairness, there is typically a dip in accuracy) [ 31 ], the random forest model with the highest observed accuracy was selected as the baseline model for further inspection of fairness.…”

“…The initial obtained data set was imbalanced (ie, there was not enough data for one class), which is a common problem in the fairness literature [ 31 ]. To mitigate the effect of imbalance, we applied the synthetic minority oversampling technique [ 32 ] to the training data (the test data remained in the original ratio).…”

“…Following the previous literature [ 36 , 37 ], men were considered the privileged group, and women were considered the unprivileged group. As there are multiple metrics to characterize accuracy in traditional ML (eg, observed accuracy, precision, recall, and F 1 score), past literature has discussed the need for multiple metrics to characterize bias in ML [ 13 , 31 ]. In this study, we adopted the five most commonly used metrics [ 15 , 16 , 38 ]:…”

Pitfalls of Machine Learning-Based Personnel Selection

“…A closely related concern is the fairness of automated personnel selection. The term algorithmic fairness (or fairness in the context of ML) means that the results of an ML algorithm are independent of certain sensitive variables (such as gender, sexuality, or religion) or proxy variables that are strongly related to them (such as zip code that can be related to race in some areas; e.g., Kleinberg et al, 2018; Pessach & Shmueli, 2020). Hence, algorithmic fairness should prevent discrimination and is highly relevant, not only for automating personnel selection procedures but also in contexts like law enforcement, healthcare, or education.…”

“…Hence, algorithmic fairness should prevent discrimination and is highly relevant, not only for automating personnel selection procedures but also in contexts like law enforcement, healthcare, or education. Pessach and Shmueli (2020) provide an overview of various measures of fairness in ML. Often these measures in some circumstances cannot be optimized simultaneously.…”