NeurIPS 2019
Sun Dec 8th through Sat the 14th, 2019 at Vancouver Convention Center
Paper ID:742
Title:Envy-Free Classification

The main points in favor of this paper are its elegant formulation of envy-freeness as a way to expand the literature on fairness (the idea being that a process is fair if no one envies the decision for any one else) The reviewers acknowledged that the connection to algorithmic fairness and the broader literature on fairness definitions was weak, but appreciated the technical development of the paper itself notwithstanding. The paper not only introduces the new notion but makes good strides on the topic of how to use it - how to construct classifiers that respect this notion and and how they might generalize.