Subset Selection Based On Multiple Rankings in the Presence of Bias: Effectiveness of Fairness Constraints for Multiwinner Voting Score Functions
Jul 24, 2023
Speakers
Niclas Boehmer
Speaker · 0 followers
L. Elisa Celis
Speaker · 0 followers
Lingxiao Huang
Speaker · 0 followers
Anay Mehrotra
Speaker · 0 followers
Nisheeth K. Vishnoi
Speaker · 0 followers
About
We consider the problem of subset selection where one is given multiple rankings of items and the goal is to select the highest "quality" subset. Score functions from the multiwinner voting literature have been used to aggregate rankings into quality scores for subsets. We study this setting of subset selection problems when, in addition, rankings may contain systemic or unconscious biases toward a group of items. For a general model of input rankings and biases, we show that requiring the selected subset to satisfy group fairness constraints can improve the quality of the selection with respect to unbiased rankings. Importantly, we show that for fairness constraints to be effective, different multiwinner score functions may require a drastically different number of rankings: While for some functions, fairness constraints need an exponential number of rankings to recover a close-to-optimal solution, for others, this dependency is only polynomial. This result relies on a novel notion of "smoothness" of submodular functions in this setting that quantifies how well a function can "correctly" assess the quality of items in the presence of bias. The results in this paper can be used to guide the choice of multiwinner score functions for the subset selection setting considered here; we additionally provide a tool to empirically enable this.We consider the problem of subset selection where one is given multiple rankings of items and the goal is to select the highest "quality" subset. Score functions from the multiwinner voting literature have been used to aggregate rankings into quality scores for subsets. We study this setting of subset selection problems when, in addition, rankings may contain systemic or unconscious biases toward a group of items. For a general model of input rankings and biases, we show that requiring the selec…
Like the format? Trust SlidesLive to capture your next event!
Professional recording and live streaming, delivered globally.
Sharing
Recommended Videos
Presentations on similar topic, category or speaker
Spurious Valleys and Clustering Behavior of Neural Networks
Watch later
Rethink DARTS Search Space and Renovate a New Benchmark
Watch later
Arithmetic Sampling: Parallel Diverse Decoding for Large Language Models
Watch later
Luke Vilnis, …
Constrained Efficient Global Optimization of Expensive Black-box Functions
Watch later
Wenjie Xu, …