A comparative study of fairness-enhancing interventions in machine learning

By Sorelle A. Friedler, Carlos Eduardo Scheidegger, Suresh Venkatasubramanian, Sonam Choudhary, Evan P. Hamilton, Derek Roth. 2018

733 citations recorded by the source corpus.

Computers are increasingly used to make decisions that have significant impact on people's lives. Often, these predictions can affect different population subgroups disproportionately. As a result, the issue of fairness has received much recent interest, and a number of fairness-enhanced classifiers have appeared in the literature. This paper seeks to study the following questions: how do these different techniques fundamentally compare to one another, and what accounts for the differences? Specifically, we seek to bring attention to many under-appreciated aspects of such fairness-enhancing interventions that require investigation for these algorithms to receive broad adoption. We present the results of an open benchmark we have developed that lets us compare a number of different algorithms under a variety of fairness measures and existing datasets. We find that although different algorithms tend to prefer specific formulations of fairness preservations, many of these measures strongly correlate with one another. In addition, we find that fairness-preserving algorithms tend to be sensitive to fluctuations in dataset composition (simulated in our benchmark by varying training-test splits) and to different forms of preprocessing, indicating that fairness interventions might be more brittle than previously thought.

Subfields: Artificial Intelligence, 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 Safety Research

In Safety, Robustness & Interpretability

Citations in the corpus

Measurement and Fairness (2019, background): “(This problem is not just theoretical pedantry; empirically, we can directly observe how these fairness-preserving algorithms fail to be reliable across variations of operationalizations and observed data [23].”
On The Impact of Machine Learning Randomness on Group Fairness (2023, background): “There has been a growing awareness of high variance in fair deep learning, associated with non-determinism in model training or the underlying implementation [3, 20, 41, 46, 49], and the uncertainty of existing results in the literature.”
Prediction-Based Decisions and Fairness: A Catalogue of Choices, Assumptions, and Definitions (2018, background): “Fairness in ML has been explored in popular books [101, 37] and a White House report [99], surveyed in technical review papers [13, 41, 19], tutorials [9, 94, 92, 23], and an in-progress textbook [10], and inspired a number of software packages [129, 39, 45].”
Stable and Fair Classification (2019, methodology): “, 2017a) since it is reported to achieve the better fairness than, and comparable accuracy to, other algorithms (Friedler et al., 2019).”
AI Fairness 360: An Extensible Toolkit for Detecting, Understanding, and Mitigating Unwanted Algorithmic Bias (2018, background): “Fairness Comparison (Friedler et al., 2018) is one of the more extensive libraries.”
Learning Fair Rule Lists (2019, background): “With respect to fairness, a significant amount of work has been done in recent years to design fairness-aware machine learning models [12, 25], which we will review in Section 2.2.”
Algorithmic Fairness (2020, methodology): “In contrast to other recent surveys in this field [37, 57], our paper proposes a comprehensive and up-to-date overview of the field, ranging from definitions and measures of fairness to state-of-the-art fairness-enhancing mechanisms.”
A Framework for Understanding Unintended Consequences of Machine Learning (2019, background): “For more analysis, Friedler et al. (2019) compare several mitigation techniques on a number of benchmark datasets.”
Bridging the Gap Between Ethics and Practice (2020, background): “Beyond detecting biases, researchers offer fairness-enhancing interventions [38] and companies have developed commercial grade toolkits for detecting and mitigating algorithmic bias [7].”
Why Is My Classifier Discriminatory? (2018, background): “For an empirical comparison, see for example Friedler et al. (2018).”
A Review of Domain Adaptation without Target Labels (2019, background): “Fairness-aware machine learning focuses on ensuring that algorithms and automated decision-making systems do not discriminate based on gender, race or other protected attributes [66], [67].”
It's COMPASlicated: The Messy Relationship between RAI Datasets and Algorithmic Fairness Benchmarks (2021, background): “[49] show that encoding choices have implications for the resulting fairness reported.”
Obtaining Fairness using Optimal Transport Theory (2018, background): “…over the Internet, they are now widely used in a large number of very sensitive areas such as medicine, human ressources with hiring policies, banks and insurance (lending), police, and justice with criminal sentencing, see for instance in [BHJ + 17] [PRT12] or [FSV + 18] and references therein.”
A Framework for Understanding Sources of Harm throughout the Machine Learning Life Cycle (2019, background): “[20] for a comparison of several of these techniques on a number of different datasets.”
Disentangling Influence: Using Disentangled Representations to Audit Model Predictions (2019, background): “, allow a fairness audit on both race and gender to be performed (without using a single combined race and gender feature [9]).”