How to be Fair and Diverse?

Abstract

Due to the recent cases of algorithmic bias in datadriven decision-making, machine learning methods are being put under the microscope in order to understand the root cause of these biases and how to correct them. Here, we consider a basic algorithmic task that is central in machine learning: subsampling from a large data set. Subsamples are used both as an end-goal in data summarization (where fairness could either be a legal, political or moral requirement) and to train algorithms (where biases in the samples are often a source of bias in the resulting model). Consequently, there is a growing effort to modify either the subsampling methods or the algorithms themselves in order to ensure fairness. However, in doing so, a question that seems to be overlooked is whether it is possible to produce fair subsamples that are also adequately representative of the feature space of the data set -an important and classic requirement in machine learning. Can diversity and fairness be simultaneously ensured? We start by noting that, in some applications, guaranteeing one does not necessarily guarantee the other, and a new approach is required. Subsequently, we present an algorithmic framework which allows us to produce both fair and diverse samples. Our experimental results on an image summarization task show marked improvements in fairness without compromising feature diversity by much, giving us the best of both the worlds.

Finding Fair and Efficient Allocations When Valuations Don't Add Up

“…Finally, the analysis of submodular valuations ties in with existing works on diversity in various elds from biology to machine learning (see, e.g. Celis et al [2016], Jost [2006]). A popular measurement for how diverse a solution is is to apply one of several concave functions called diversity indices to the proportions of the di erent entities/a ributes (with respect to which we wish to be diverse) in the solution, e.g.…”

Algorithm Fairness in AI for Medicine and Healthcare

“…As a result, the development of frameworks for understanding data subsampling has emerged as its own subfield tangent to fair machine learning. One such framework is determinantal point process (k-DPP), which proposes quantifiable measures for combinatorial subgroup diversity (via Shannon Entropy) and geometric feature diversity (via measuring the volume of the k-dimensional feature space) 93 . In tandem with resampling is the problem of optimal data (or resource) allocation in operationalizing dataset collection in statistical surveys, as well as game-theoretic frameworks for understanding the influence of individual data points via Shapley values, which may refine resampling techniques to developing fair and diverse training datasets [94][95][96][97] .…”

Socially Responsible AI Algorithms: Issues, Purposes, and Challenges

“…Fair machine learning algorithms need to adopt/create specific fairness definitions that fit into context [87,111,119,143,31]. Common methods in fair classification include blinding [32,93], causal methods [80,108], transformation [74,64,208], sampling and subgroup analysis [37,65], adversarial learning [75,211,175], reweighing [106,31], and regularization and constraint optimization [111,17].…”