The Myth in the Methodology: Towards a Recontextualization of Fairness in Machine Learning

By Ben Green, Lily Hu. 2018

108 citations recorded by the source corpus.

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Citations in the corpus

Prediction-Based Decisions and Fairness: A Catalogue of Choices, Assumptions, and Definitions (2018, background): “Both can narrow focus to the quantifiable, introduce harmful simplifications, and mislead that the issues are purely technical [50, 96].”
Improving Fairness in Machine Learning Systems: What Do Industry Practitioners Need? (2018, background): “ewees highlighted limitations of existing UX prototyping methods for surfacing fairness issues in complex, multi-component ML systems (e.g., chatbots), where fairness can be heavily context dependent [46, 71], and the space of potential contexts is often very large. The rapidly growing area of fairn”
Towards a critical race methodology in algorithmic fairness (2019, neutral): “libration (or test fairness) [23,66]requirestheactualoutcometobeindependentofprotected attributes, conditioned on estimated outcome. Whilelimitationsofgroupfairnesscriteriahavebeenpreviously explored [26, 47],here we focus ourexamination on theconceptualization of race embedded within this framework”
It's COMPASlicated: The Messy Relationship between RAI Datasets and Algorithmic Fairness Benchmarks (2021, background): “Fairness as a social ideal is unstable and uncertain, constantly being debated and shaped, "a process of continual social negotiation and adjudication between competing needs and visions of the good" [75].”
The false promise of risk assessments: epistemic reform and the limits of fairness (2020, background): “abstract, apart from social, political, and historical context [63].”
A Framework for Understanding Unintended Consequences of Machine Learning (2019, background): “This assumption often does not address domain-specific societal and historical contexts (Green and Hu, 2018).”
Algorithmic realism: expanding the boundaries of algorithmic thought (2020, background): “Even within the FAT* community, critical evaluations of the mathematization of fairness suggest the need for further evolution from formalism towards realism [11, 52, 58, 67, 135].”
Algorithmic Bias in Education (2021, background): “Work by Green and colleagues (Green, 2020; Green & Hu, 2018; Green & Viljoen, 2020), for example, suggests that the conflation of fairness defined as statistical measures of group parity with fairness in society creates its own obstacles to addressing social inequities.”
Why Machine Learning May Lead to Unfairness: Evidence from Risk Assessment for Juvenile Justice in Catalonia (2019, background): “However, the fact that base rates differ so substantially can already be the result of years of structural discrimination long before the present data was recorded [22].”
REVISE: A Tool for Measuring and Mitigating Bias in Visual Datasets (2020, background): “Further, dataset bias mitigation is only one step, albeit an important one, in the much broader process of addressing fairness in the deployment of a machine learning sys-tem (Birhane, 2021; Green and Hu, 2018).”
A Review on Fairness in Machine Learning (2022, background): “Green and Hu [88] note that this definition of similarity should be based on the social context and judgments of humans, not solely on mathematical measurements.”
Disparate Interactions: An Algorithm-in-the-Loop Analysis of Fairness in Risk Assessments (2019, methodology): “We focused on calibration not as an ideal metric for fairness (recognizing that no perfect metric for fairness can exist [32]), but because it is the most commonly-used approach for evaluating risk assessments in practice [16, 25, 44].”
Towards Intersectionality in Machine Learning: Including More Identities, Handling Underrepresentation, and Performing Evaluation (2022, background): “We also note that despite the language we employ, we do not suggest that fairness can be treated as a purely algorithmic problem that neglects the sociotechnical frame [12, 50, 51, 109].”
Algorithmic Fairness: Choices, Assumptions, and Definitions (2021, background): “Several recent papers have demonstrated how policy goals are, sometimes clumsily, abstracted and formulated to fit into a prediction task (Dobbe et al. 2018, Eckhouse et al. 2018, Green 2018, Green & Hu 2018, Ochigame et al. 2018, Silva & Kenney 2018, Passi & Barocas 2019, Selbst et al. 2019).”