Poster
in
Workshop: Algorithmic Fairness through the lens of Metrics and Evaluation
Improving Fairness in Matching under Uncertainty
Piyushi Manupriya
Keywords: [ Algorithm Development ] [ Fairness ]
The two-sided matching problem is prevalent in numerous applications, such as matching employers with employees, students with universities, etc. An emerging concern is that of incorporating fairness in matching that seeks to satisfy the preferences of the entities involved in addition to maximizing the utility obtained. Such problems become more challenging when preferences are uncertain due to confounding factors present in the real world. For e.g., universities often prefer students based on their grades which could be an incomplete indicator of the student's merit. Recent work of Devic et al. [2023] presents an axiomatic approach to this problem of fairness in matching under uncertainty. However, a notable drawback of the fairness axioms in Devic et al. [2023] is that their fairness guarantees are limited to only one of the sides. Our work proposes an improved axiomatic approach to this problem with fairness to both sides and computational efficiency.