Fairness in Machine Learning

By Luca Oneto, Silvia Chiappa. 2020

557 citations recorded by the source corpus.

Machine learning based systems are reaching society at large and in many aspects of everyday life. This phenomenon has been accompanied by concerns about the ethical issues that may arise from the adoption of these technologies. ML fairness is a recently established area of machine learning that studies how to ensure that biases in the data and model inaccuracies do not lead to models that treat individuals unfavorably on the basis of characteristics such as e.g. race, gender, disabilities, and sexual or political orientation. In this manuscript, we discuss some of the limitations present in the current reasoning about fairness and in methods that deal with it, and describe some work done by the authors to address them. More specifically, we show how causal Bayesian networks can play an important role to reason about and deal with fairness, especially in complex unfairness scenarios. We describe how optimal transport theory can be leveraged to develop methods that impose constraints on the full shapes of distributions corresponding to different sensitive attributes, overcoming the limitation of most approaches that approximate fairness desiderata by imposing constraints on the lower order moments or other functions of those distributions. We present a unified framework that encompasses methods that can deal with different settings and fairness criteria, and that enjoys strong theoretical guarantees. We introduce an approach to learn fair representations that can generalize to unseen tasks. Finally, we describe a technique that accounts for legal restrictions about the use of sensitive attributes.

Subfields: Artificial Intelligence, Safety Research

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

Citations in the corpus

50 Years of Test (Un)fairness: Lessons for Machine Learning (2018, background): “• Peterson and Novick’s “conditional probability and its con-verse” sufficiency [4], equalized odds [32], or conditional procedure accuracy [5], sometimes expressed as the conditional independence A ⊥ D | Y .”
Challenges in Deploying Machine Learning: A Survey of Case Studies (2020, background): “Such questions about fairness and accountability of ML algorithms and data are studied by the branch of machine learning known as Fairness in ML [16].”
This Thing Called Fairness (2019, background): “7For an overview of many available statistical fairness measures, see several tutorials at major machine learning research conferences [13, 57, 108] as well as excellent summaries in [103, 104, 148].”
FUTURE-AI: international consensus guideline for trustworthy and deployable artificial intelligence in healthcare (2023, unknown): “When possible—that is, the individuals’ and data attributes are available—bias detection methods should be applied by using fairness metrics such as true positive rates, statistical parity, group fairness, and equalised odds.”
Algorithm fairness in artificial intelligence for medicine and healthcare (2023, background): “However, as emphasized previously 391 , who bears the cost of misclassification?”
Economics, Fairness and Algorithmic Bias (2020, unknown): “55 Importantly, many papers contain impossibility results showing that many attractive fairness criteria cannot be simultaneously achieved (Kleinberg et al., 2016; Barocas et al., 2017; Chouldechova, 2017; Berk et al., 2017; Narayanan, 2018c).”
NOTE: Robust Continual Test-time Adaptation Against Temporal Correlation (2022, background): “As TTA relies on unlabeled test samples and changes the model accordingly, the model is exposed to potential data-driven biases after adaptation, such as fairness issues [3] and adversarial attacks [36].”
Stable Bias: Evaluating Societal Representations in Diffusion Models (2023, unknown): “…previous work in ML has adopted binary gender and fixed prior ethnicity categorizations for the sake of convenience, the downstream impact of these choices that are propagated from (labeled) datasets to trained models can contribute to perpetuating algorithmic harms and unfairness [7, 5, 31, 15].”