Diverse Group Formation Based on Multiple Demographic Features

TL;DR

This paper introduces a novel approach for forming diverse groups by considering multiple demographic features simultaneously, aiming to enhance fairness and diversity in team formation algorithms.

Contribution

It proposes a new method to model demographic profiles and two diversity ranking algorithms that incorporate multiple demographic attributes into team formation.

Findings

Algorithms produce more diverse groups with acceptable utility loss.

Multidimensional demographic features improve diversity in team formation.

Evaluation on real dataset demonstrates effectiveness of the proposed methods.

Abstract

The goal of group formation is to build a team to accomplish a specific task. Algorithms are employed to improve the effectiveness of the team so formed and the efficiency of the group selection process. However, there is concern that team formation algorithms could be biased against minorities due to the algorithms themselves or the data on which they are trained. Hence, it is essential to build fair team formation systems that incorporate demographic information into the process of building the group. Although there has been extensive work on modeling individuals expertise for expert recommendation and or team formation, there has been relatively little prior work on modeling demographics and incorporating demographics into the group formation process. We propose a novel method to represent experts demographic profiles based on multidimensional demographic features. Moreover, we introduce two diversity ranking algorithms that form a group by considering demographic features along with the minimum required skills. Unlike many ranking algorithms that consider one Boolean demographic feature (e.g., gender or race), our diversity ranking algorithms consider multiple multivalued demographic attributes simultaneously. We evaluate our proposed algorithms using a real dataset based on members of a computer science program committee. The result shows that our algorithms form a program committee that is more diverse with an acceptable loss in utility.