CoChat
Sign Up
Cited by 1,124Ethics and Social Impacts of AIExplainable Artificial Intelligence (XAI)Discrimination and Equality Law

Fairness definitions explained

Verma, Sahil, Rubin, Julia·2018·Proceedings of the International Workshop on Software Fairness
DOI 10.1145/3194770.3194776
Open this paper in CoChat — free account

Abstract

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.

Integrity signals

Registry

Crossref · 10.1145/3194770.3194776

Cites

20 of 20 references verified0 citation errors · 0 not checkable

Statistics

Recomputing p-values needs the paper’s full text.

Claims the paper makes

This paper’s full text isn’t openly available, so its claims and self-check weren’t run. The registry and citation checks still apply.

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

  • Parity on different metrics across disaggregated population subgroups corresponds to how fairness is often defined.

    “Parity on the different metrics across disaggregated population subgroups corresponds to how fairness is often defined [37, 48].”

    From Model Cards for Model Reporting (2019) · cites this paper as [48]

  • Less has been done in the NLP space on fairness desiderata than in machine learning classification tasks, and NLP may contain more ambiguity than classification for tasks like language generation.

    “While similar concepts have been operationalized for machine learning classification tasks (Mehrabi et al. 2021; Verma and Rubin 2018), less has been done in the NLP space, which may contain more ambiguity than classification for tasks like language generation.”

    From Bias and Fairness in Large Language Models: A Survey (2024) · cites this paper as Verma and Rubin 2018

  • Strategies such as robust data augmentation, counterfactual fairness, and diverse representative datasets alongside unbiased data collection methods can combat bias.

    “To combat bias, this paper has highlighted strategies such as robust data augmentation, the application of counterfactual fairness, and the imperative for diverse, representative datasets alongside unbiased data collection methods [49][50][51].”

    From Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies (2023) · cites this paper as [51]

  • Scholars have proposed a number of competing ways to quantify fairness, some of which are mathematically incompatible with each other.

    “As we discussed in §2.3 scholars have proposed a number of competing ways to quantify fairness [46,75,93,148], some of which are mathematically incompatible with each other [26,51,80].”

    From Out of Context: Investigating the Bias and Fairness Concerns of “Artificial Intelligence as a Service” (2023) · cites this paper as [148]

  • A classifier trained without regard to the protected attribute fails to remove discrimination and can complicate detection and mitigation of discrimination downstream via in-processing or post-processing techniques.

    “As a result, a classifier trained without regard to the protected attribute not only fails to remove discrimination, but it can complicate the detection and mitigation of discrimination downstream via in-processing or post-processing techniques [9,10,18,24,25,34,43,51], such as those we describe next.”

    From Interventional Fairness: Causal Database Repair for Algorithmic Fairness (2019) · cites this paper as [51]

Cites

Flagged first, then by citations

1 of the 21 references couldn’t be matched to a known paper.

Cited by

Sorted by citations

Related

Authors
Verma, S., & Rubin, J.
Affiliations
  • Indian Institute of Technology Kanpur (IN)
  • University of British Columbia (CA)
ORCID
Venue
Proceedings of the International Workshop on Software Fairness
Publisher
ACM
Published
2018-05-29
Pages
1-7
DOI
10.1145/3194770.3194776
OpenAlex
W2809878087
Cited by
1,124
References
21
Citations by year
2026: 102 · 2025: 157 · 2024: 188 · 2023: 197 · 2022: 168 · 2021: 170 · 2020: 103 · 2019: 38 · 2018: 1
FWCI
140.65
Type
conference paper
Language
en
Open access
closed
Licence
www.acm.org/publications/policies/copyright_policy#Background (vor)
Topics
  • Social Sciences › Safety Research › Ethics and Social Impacts of AI
  • Physical Sciences › Computer Science › Artificial Intelligence › Explainable Artificial Intelligence (XAI)
  • Social Sciences › Law › Discrimination and Equality Law
Keywords
algorithmic fairness, algorithmic classification, fairness metrics

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review