Price of anarchy and welfare loss in other algorithmic monoculture models

Determine how detrimental algorithmic monoculture is, in terms of the price of anarchy or social welfare loss, in the two-sided matching model studied by Peng and Garg and in the competing-classifiers model studied by Jagadeesan et al., as well as in other formal models of algorithmic monoculture beyond the one-sided matching setting analyzed in this paper.

Background

This paper establishes a tight constant bound of 2 on the price of anarchy for a one-sided matching model with advice under a broad class of stochastically consistent ranking technologies, thereby quantifying the worst-case social welfare loss due to algorithmic monoculture in that setting.

While this result resolves an open question for the one-sided matching framework introduced by Kleinberg and Raghavan, analogous worst-case welfare guarantees for other models of algorithmic monoculture—such as two-sided matching markets and settings with competing classifiers—are not provided here and explicitly remain open.