Reliability at Multiple Stages in a Data Analysis Pipeline.

Saved in:

Bibliographic Details
Title:	Reliability at Multiple Stages in a Data Analysis Pipeline.
Authors:	MOSKOVITCH, YUVAL^1,2,3 yuvalmos@bgu.ac.il, JAGADISH, H. V.^4,5,6
Source:	Communications of the ACM. Nov2022, Vol. 65 Issue 11, p118-128. 11p. 2 Diagrams, 4 Charts.
Subjects:	Decision support systems, Software reliability, Racism, Sexism, Fairness, Machine learning, Algorithms
Abstract:	This article details how to incorporate reliability in data-driven decision-making tools. The importance of reliability in this type of software is explained, involving the affect on people’s day-to-day lives and the existing evidence of racial and gender bias in these tools. Ways to improve reliability in the tools is discussed, including appropriate datasets, pattern count-based labels, label computation and fairness measures. The author focuses on categorical data and details various angles where reliability and fairness can be examined in these decision-making tools.
Database:	Engineering Source

Description
Abstract:	This article details how to incorporate reliability in data-driven decision-making tools. The importance of reliability in this type of software is explained, involving the affect on people’s day-to-day lives and the existing evidence of racial and gender bias in these tools. Ways to improve reliability in the tools is discussed, including appropriate datasets, pattern count-based labels, label computation and fairness measures. The author focuses on categorical data and details various angles where reliability and fairness can be examined in these decision-making tools.
ISSN:	00010782
DOI:	10.1145/3500923