Computers and algorithms are increasingly pervading our daily lives, therefore to trust these systems we have to make sure that the data they use are fair and without bias. As a result, Fairness has become a relevant topic of discussion within the field of Data Science, and technologies that accurately discover discrimination and bias present in datasets are of paramount importance. In this work we present FAIR-DB (FunctionAl dependencIes to discoveR Data Bias), a novel framework to detect biases and discover discrimination in datasets. By exploiting various kinds of functional dependencies, our tool can identify those attributes in a database that encompass discrimination (e.g. gender, ethnicity or religion) and the ones that instead satisfy various fairness criteria. We compared our framework with two state-of-the-art systems for detecting unfairness in datasets, obtaining overall similar results on a real-world dataset; specifically, the comparison highlighted that FAIR-DB not only provides very precise information about the groups treated unequally, but also that, in comparison with other existing tools, may obtain more insights regarding the bias present in datasets.

A short account of FAIR-DB: A system to discover Data Bias

Azzalini F.;Tanca L.
2021

Abstract

Computers and algorithms are increasingly pervading our daily lives, therefore to trust these systems we have to make sure that the data they use are fair and without bias. As a result, Fairness has become a relevant topic of discussion within the field of Data Science, and technologies that accurately discover discrimination and bias present in datasets are of paramount importance. In this work we present FAIR-DB (FunctionAl dependencIes to discoveR Data Bias), a novel framework to detect biases and discover discrimination in datasets. By exploiting various kinds of functional dependencies, our tool can identify those attributes in a database that encompass discrimination (e.g. gender, ethnicity or religion) and the ones that instead satisfy various fairness criteria. We compared our framework with two state-of-the-art systems for detecting unfairness in datasets, obtaining overall similar results on a real-world dataset; specifically, the comparison highlighted that FAIR-DB not only provides very precise information about the groups treated unequally, but also that, in comparison with other existing tools, may obtain more insights regarding the bias present in datasets.
CEUR Workshop Proceedings
File in questo prodotto:
Non ci sono file associati a questo prodotto.

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: http://hdl.handle.net/11311/1207990
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 1
  • ???jsp.display-item.citation.isi??? ND
social impact