Explaining contributions of features towards unfairness in classifiers: A novel threshold-dependent Shapley value-based approach

Página do item simplificado
dc.contributor.authorPelegrina G.D.
dc.contributor.authorSiraj S.
dc.contributor.authorDuarte L.T.
dc.contributor.authorGrabisch M.
dc.date.accessioned2024-11-01T06:11:44Z
dc.date.available2024-11-01T06:11:44Z
dc.date.issued2024
dc.description.abstract© 2024 Elsevier LtdA number of approaches has been proposed to investigate and mitigate unfairness in machine learning algorithms. However, as the definition and understanding of fairness may vary in different situations, the study of ethical disparities remains an open area of research. Besides the importance of analyzing ethical disparities, explainability in machine learning is also a relevant issue in Trustworthy Artificial Intelligence. Usually, both fairness and explainability analysis are based on a fixed decision threshold, which differentiates the positive cases from the negative ones according to the predicted probabilities. In this paper, we investigate how changes in this threshold can impact the fairness of predictions between protected and other groups and how features contribute towards such a measure. We propose a novel Shapley value-based approach as a tool to investigate how changes in the threshold values change the contribution of each feature towards unfairness. This gives us an ability to evaluate how fairness measures vary for different threshold values and which features have the higher (or lower) impact on creating ethical disparities. We demonstrate this using three different case studies that are carefully chosen to highlight different unfairness scenarios and features contributions. We also applied our proposal as a feature selection strategy, which contributed to decrease unfair results substantially.
dc.description.volume138
dc.identifier.doi10.1016/j.engappai.2024.109427
dc.identifier.issnNone
dc.identifier.urihttps://dspace.mackenzie.br/handle/10899/39676
dc.relation.ispartofEngineering Applications of Artificial Intelligence
dc.rightsAcesso Restrito
dc.subject.otherlanguageFairness
dc.subject.otherlanguageFeature contribution
dc.subject.otherlanguageInterpretable machine learning
dc.subject.otherlanguageShapley value
dc.titleExplaining contributions of features towards unfairness in classifiers: A novel threshold-dependent Shapley value-based approach
dc.typeArtigo
local.scopus.citations0
local.scopus.eid2-s2.0-85205447975
local.scopus.subjectDecision threshold
local.scopus.subjectFairness
local.scopus.subjectFairness measures
local.scopus.subjectFeature contribution
local.scopus.subjectInterpretable machine learning
local.scopus.subjectMachine learning algorithms
local.scopus.subjectMachine-learning
local.scopus.subjectShapley value
local.scopus.subjectThreshold-value
local.scopus.subjectValue-based approach
local.scopus.updated2024-12-01
local.scopus.urlhttps://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85205447975&origin=inward
Arquivos
Coleções
Artigos de periódico