AF:RI:Small: Fairness in allocation and machine learning problems: algorithms and solution concepts

点击翻译按钮获取中文摘要

Fairness is one of the highest pursuits of human society. With advanced computer technology, enormous computational power, and data availability, it is natural and inevitable to turn to computer-aided decision-making. There arises a two-fold challenge of fair computation: (i) achieve fairness through computation, i.e., allocating resources (or tasks) fairly, for example, in vaccine distribution and sharing pollution costs, and (ii) fairness in computation arising from machine learning-based decision making, for example, justice decisions and deciding medical treatment. Centering around the fundamental solution concepts from economics and social choice theory, this project will significantly advance the state-of-the-art on both challenges. In the process, it will develop fair solutions of societal importance and build theory at the intersection of economics, machine learning, operations research, social choice theory, and theoretical computer science. In addition, the proposed educational activities will create many unique research opportunities for students at all levels, from high school to graduate students, generating high-quality researchers and practitioners for society.This project will (i) tackle the most important open problems on the existence and computation of classical fair solutions and (ii) develop preference-based fairness concepts for the machine learning tasks by marrying them to the rich literature on social choice theory. The former may provide radical breakthroughs on some of the most enigmatic open questions in the field. At the same time, the latter may pave the way for an alternate theory of fair machine learning that is more nuanced and is rooted in classical concepts from social choice theory.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.