Fair and Diverse DPP-based Data Summarization

By L. Elisa Celis, Vijay Keswani, Damian Straszak, Amit Deshpande, Tarun Kathuria, Nisheeth K. Vishnoi. 2018

135 citations recorded by the source corpus.

Sampling methods that choose a subset of the data proportional to its diversity in the feature space are popular for data summarization. However, recent studies have noted the occurrence of bias (under- or over-representation of a certain gender or race) in such data summarization methods. In this paper we initiate a study of the problem of outputting a diverse and fair summary of a given dataset. We work with a well-studied determinantal measure of diversity and corresponding distributions (DPPs) and present a framework that allows us to incorporate a general class of fairness constraints into such distributions. Coming up with efficient algorithms to sample from these constrained determinantal distributions, however, suffers from a complexity barrier and we present a fast sampler that is provably good when the input vectors satisfy a natural property. Our experimental results on a real-world and an image dataset show that the diversity of the samples produced by adding fairness constraints is not too far from the unconstrained case, and we also provide a theoretical explanation of it.

Subfields: Artificial Intelligence

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Lingxiao Huang
2. L. Elisa Celis
3. Vijay Keswani
4. Nisheeth K. Vishnoi
5. Prasad Raghavendra

Citations in the corpus

• Stable and Fair Classification (2019, background): “, ranking [18, 76], summarization [17], personalization [19, 16], multiwinner voting [15], and online advertising [13].”
• Multiwinner Voting with Fairness Constraints (2017, background): “More generally, our results contribute to the growing set of algorithms that incorporate fairness constraints to counter algorithmic bias in fundamental algorithmic problems such as classification [20, 54, 52, 51], sampling [13, 12, 14], ranking [15, 50] and personalization [16].”
• Classification with Fairness Constraints: A Meta-Algorithm with Provable Guarantees (2018, methodology): “To handle this problem, we consider linear fairness constraints, which have been considered in other fundamental algorithmic problems including sampling [11, 10, 13], ranking [14], voting [12], and personalization [15].”
• Implicit Diversity in Image Summarization (2019, background): “Many existing fair summarization algorithms assume the presence of socially salient attribute labels to generate fair summaries [56, 19] by using the labels to enforce fairness constraints on the output summary.”
• Toward Controlling Discrimination in Online Ad Auctions (2019, background): “Several prior works address the problems of polarization and algorithmic bias, including [10,24] who control polarization in social-networks and personalized feeds, [42] who diversify personal feeds, and [12] who create a diverse and balanced summary of a set of results.”
• Dialect Diversity in Text Summarization on Twitter (2020, background): “For example, [12, 40] use labels to construct fairness constraints or scoring functions to guarantee appropriate diversity in automatically generated summaries.”