Addressing Bias in Algorithmic Solutions: Exploring Vertex Cover and Feedback Vertex Set

TL;DR

This paper investigates how to incorporate fairness considerations into combinatorial optimization problems like Vertex Cover and Feedback Vertex Set, aiming to find solutions that are unbiased across different subgroups.

Contribution

It introduces a framework for defining and computing unbiased solutions in combinatorial optimization, addressing the impact of subgroup disparities.

Findings

Proposes methods to incorporate subgroup fairness into optimization algorithms

Demonstrates that unbiased solutions can be effectively computed

Highlights the importance of fairness in real-world optimization problems

Abstract

A typical goal of research in combinatorial optimization is to come up with fast algorithms that find optimal solutions to a computational problem. The process that takes a real-world problem and extracts a clean mathematical abstraction of it often throws out a lot of "side information" which is deemed irrelevant. However, the discarded information could be of real significance to the end-user of the algorithm's output. All solutions of the same cost are not necessarily of equal impact in the real-world; some solutions may be much more desirable than others, even at the expense of additional increase in cost. If the impact, positive or negative, is mostly felt by some specific (minority) subgroups of the population, the population at large will be largely unaware of it. In this work we ask the question of finding solutions to combinatorial optimization problems that are "unbiased" with respect to a collection of specified subgroups of the total population.