Fairness in Machine Learning: A Survey
Справедливость в машинном обучении: обзор
2023-08-23
SCID: 54.1/s5fs2t6w
Discuss with AI
algorithmic bias mitigationfairness in recommender systemsfairness in regressionmachine learning fairnesspre-processing, in-processing, and post-processing
Figures from the paper
Abstract (AI)
When Machine Learning technologies are used in contexts that affect citizens, companies as well as researchers need to be confident that there will not be any unexpected social implications, such as bias towards gender, ethnicity, and/or people with disabilities. There is significant literature on approaches to mitigate bias and promote fairness, yet the area is complex and hard to penetrate for newcomers to the domain. This article seeks to provide an overview of the different schools of thought and approaches that aim to increase the fairness of Machine Learning. It organizes approaches into the widely accepted framework of pre-processing, in-processing, and post-processing methods, subcategorizing into a further 11 method areas. Although much of the literature emphasizes binary classification, a discussion of fairness in regression, recommender systems, and unsupervised learning is also provided along with a selection of currently available open source libraries. The article concludes by summarizing open challenges articulated as five dilemmas for fairness research.
Key Findings
1
Beyond binary classification, the survey examines fairness in regression, recommender systems, and unsupervised learning.
2
It concludes by identifying five dilemmas that define major open challenges for fairness research.
3
It synthesizes diverse approaches for mitigating bias related to gender, ethnicity, disability, and other socially sensitive attributes.
4
The article reviews currently available open-source libraries supporting fairness research and implementation.
5
The survey organizes machine-learning fairness interventions into pre-processing, in-processing, and post-processing frameworks, further divided into 11 method areas.
Research Object
Machine learning technologies and models used in socially consequential contexts
Research Subject
Fairness and bias mitigation across pre-processing, in-processing, and post-processing approaches, including effects related to gender, ethnicity, and disability
Publication Details
Publication Date
2023-08-23
Journal
Publisher
ISSN
Cited by
991
Open access PDF
Access Type
Author Information
Download PDF
Subscribe to digest