A Survey of Bias in Machine Learning Through the Prism of Statistical Parity

Besse, Philippe; Barrio, Eustasio del; Gordaliza, Paula; Loubes, Jean–Michel; Risser, Laurent

doi:10.1080/00031305.2021.1952897

Cited by 27 publications

(20 citation statements)

References 29 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…That is, each category has an equal probability of being classified as positive. However, a downside of this notion is that it ignores possible variations between classes [14].…”

Section: Bias and Fairness Measurement Methods A) Metrics Focused On ...mentioning

confidence: 99%

Monitoring bias and fairness in machine learning models: A review

Bacelar

2021

Preprint

View full text Add to dashboard Cite

Introduction Machine learning algorithms are quickly gaining traction in both the private and public sectors for their ability to automate both simple and complex decision-making processes. The vast majority of economic sectors, including transportation, retail, advertisement, and energy, are being disrupted by widespread data digitization and the emerging technologies that leverage it. Computerized systems are being introduced in government operations to improve accuracy and objectivity, and AI is having an impact on democracy and governance [1]. Numerous businesses are using machine learning to analyze massive quantities of data, from calculating credit for loan applications to scanning legal contracts for errors to analyzing employee interactions with customers to detect inappropriate behavior. New tools make it easier than ever for developers to design and deploy machine-learning algorithms [2] [3].

show abstract

“…That is, each category has an equal probability of being classified as positive. However, a downside of this notion is that it ignores possible variations between classes [14].…”

Section: Bias and Fairness Measurement Methods A) Metrics Focused On ...mentioning

confidence: 99%

Monitoring bias and fairness in machine learning models: A review

Bacelar

2021

Preprint

View full text Add to dashboard Cite

show abstract

“…However, the use of such algorithms requires the protected attributes to be available in the model usage stage. Besse et al (2021) provide another example of using post-processing method to mitigate demographic parity.…”

Section: De-biasing and Mitigating Unfairnessmentioning

confidence: 99%

“…Besse et al . (2021) provide another example of using post‐processing method to mitigate demographic parity.…”

Section: De‐biasing and Mitigating Unfairnessmentioning

confidence: 99%

Bias, Fairness and Accountability with Artificial Intelligence and Machine Learning Algorithms

Zhou

Zhang

Nair

et al. 2022

Int Statistical Rev

View full text Add to dashboard Cite

Summary The advent of artificial intelligence (AI) and machine learning algorithms has led to opportunities as well as challenges in their use. In this overview paper, we begin with a discussion of bias and fairness issues that arise with the use of AI techniques, with a focus on supervised machine learning algorithms. We then describe the types and sources of data bias and discuss the nature of algorithmic unfairness. In addition, we provide a review of fairness metrics in the literature, discuss their limitations, and describe de‐biasing (or mitigation) techniques in the model life cycle.

show abstract

“…• EDF completely excludes S, thus complying with European Union regulations [6]. [20] and [17] go much further, allowing limited involvement of S (even in the prediction stage), thus possibly out of compliance with legal requirements in some jurisdictions.…”

Section: Relation To the Work Of Komiyama Et Al And Scutari Et Almentioning

confidence: 99%

A Novel Regularization Approach to Fair ML

Matloff¹,

Zhang²

2022

Preprint

View full text Add to dashboard Cite

A number of methods have been introduced for the fair ML issue, most of them complex and many of them very specific to the underlying ML moethodology. Here we introduce a new approach that is simple, easily explained, and potentially applicable to a number of standard ML algorithms. Explicitly Deweighted Features (EDF) reduces the impact of each feature among the proxies of sensitive variables, allowing a different amount of deweighting applied to each such feature. The user specifies the deweighting hyperparameters, to achieve a given point in the Utility/Fairness tradeoff spectrum. We also introduce a new, simple criterion for evaluating the degree of protection afforded by any fair ML method. * NM is professor of computer science, UCD, and has served as an expert witness in various litigation cases involving discrimination.† WZ performed this work as an undergraduate at UCD.

show abstract

A Survey of Bias in Machine Learning Through the Prism of Statistical Parity

Cited by 27 publications

References 29 publications

Monitoring bias and fairness in machine learning models: A review

Monitoring bias and fairness in machine learning models: A review

Bias, Fairness and Accountability with Artificial Intelligence and Machine Learning Algorithms

A Novel Regularization Approach to Fair ML

Contact Info

Product

Resources

About