Fairness Definitions Explained

By Sahil Verma, Julia Rubin. 2018

1,328 citations recorded by the source corpus.

Algorithm fairness has started to attract the attention of researchers in AI, Software Engineering and Law communities, with more than twenty different notions of fairness proposed in the last few years. Yet, there is no clear agreement on which definition to apply in each situation. Moreover, the detailed differences between multiple definitions are difficult to grasp. To address this issue, this paper collects the most prominent definitions of fairness for the algorithmic classification problem, explains the rationale behind these definitions, and demonstrates each of them on a single unifying case-study. Our analysis intuitively explains why the same case can be considered fair according to some definitions and unfair according to others.

Subfields: Artificial Intelligence, Information Systems, Safety Research, Safety, Robustness & Interpretability

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Information Systems

In Safety Research

In Safety, Robustness & Interpretability

Citations in the corpus

Model Cards for Model Reporting (2018, background): “FalseOmissionRate, which measure the fraction of negative (not smiling) and positive (smiling) predictions that are incorrectly predicted to be positive and negative, respectively, are also reported. [48] • Together, these four metrics provide values for different errors that can be calculated from ”
FairFace: Face Attribute Dataset for Balanced Race, Gender, and Age (2019, background): “There exist many different definitions of fairness used in the literature [61].”
Assessing Social and Intersectional Biases in Contextualized Word Representations (2019, background): “Downstream tasks such as classification and ranking will suffer from the implicit encoding of these biases, posing a hurdle to building fair machine learning systems [31].”
A Survey on Bias and Fairness in Machine Learning (2019, methodology): “Here we will reiterate and provide some of the most widely used definitions, along with their explanations inspired from [122].”
Theoretical Guarantees for Model Auditing with Finite Adversaries (2019, background): “We remark that this framework is typical in the fairness literature, where a vast array of notions of fairness have been proposed based on the statistical dependency of Ŷ and Z (see, for example [34] and the references therein).”
Fair decision making using privacy-protected data (2019, background): “Fairness has been formalized in economics, and, more recently, in definitions emerging from machine learning [12, 13, 24, 27, 31].”
Prediction-Based Decisions and Fairness: A Catalogue of Choices, Assumptions, and Definitions (2018, background): “Fairness in ML has been explored in popular books [98, 34] and a White House report [35], surveyed in technical review papers [8, 40, 15, 126], and an in-progress textbook [5], and inspired a number of software packages [132, 38, 2, 43, 47].”
Social Data: Biases, Methodological Pitfalls, and Ethical Boundaries (2019, background): “Researchers have also developed a growing body of evaluation metrics for measuring fairness (Dwork et al., 2012; Kearns et al., 2017; Heidari et al., 2018; Verma and Rubin, 2018).”
Efficient candidate screening under multiple tests and implications for fairness (2019, background): “Finally, [17] provide a survey of definitions from the algorithmic fairness literature.”
Learning Fair Rule Lists (2019, background): “We refer the interested reader to the following surveys [6, 14, 50, 58] for a detailed overview of these notions.”
Algorithmic Fairness (2020, background): “For further reading about algorithmic fairness measures, we refer the reader to [38], [85], and [136].”
Counterfactual Explanations for Machine Learning: A Review (2020, background): “lso attract wide-spread criticism and are therefore important to avoid [55, 69]. Fairness has been captured in several notions, based on a demographic grouping or individual capacity. Verma and Rubin [109] have enumerated, and intuitively explained many fairness definitions using a unifying dataset.”
Bias in word embeddings (2020, background): “[42, 85, 93] • What are the actual consequences of biased algorithms and who is accountable for them?”
A Geometric Solution to Fair Representations (2020, background): “While there exists no consensus for measuring fairness, researchers have proposed a variety of metrics, some focusing on representations and some on the predicted outcomes [13, 31].”
Interventional Fairness: Causal Database Repair for Algorithmic Fairness (2019, background): “A large number of fairness definitions have been proposed (see Verma and Rubin for a recent discussion [52]), which we broadly categorize in Fig.”
Underdiagnosis bias of artificial intelligence algorithms applied to chest radiographs in under-served patient populations (2021, background): “One such definition is predictive parity, which implies equal positive predictive value, or, equivalently, false discovery rate (FDR) between the groups 30 .”
A Framework for Understanding Unintended Consequences of Machine Learning (2019, background): “Interested readers are referred to (Narayanan, 2018) or (Verma and Rubin, 2018) for a more detailed discussion of different fairness definitions.”