On Finding Stable and Efficient Solutions for the Team Formation Problem

TL;DR

This paper presents a novel mathematical programming approach to the team formation problem, balancing efficiency and stability, and introduces a branch-cut-and-price algorithm validated through extensive simulations.

Contribution

It models team formation as a bi-level optimization problem balancing efficiency and stability, and develops an effective branch-cut-and-price algorithm.

Findings

The proposed algorithm outperforms existing methods in simulations.

Balancing social welfare and stability yields more practical team formations.

The approach is applicable to various real-world team assignment scenarios.

Abstract

The assignment of personnel to teams is a fundamental and ubiquitous managerial function, typically involving several objectives and a variety of idiosyncratic practical constraints. Despite the prevalence of this task in practice, the process is seldom approached as a precise optimization problem over the reported preferences of all agents. This is due in part to the underlying computational complexity that occurs when quadratic (i.e., intra-team interpersonal) interactions are taken into consideration, and also due to game-theoretic considerations, when those taking part in the process are self-interested agents. Variants of this fundamental decision problem arise in a number of settings, including, for example, human resources and project management, military platooning, sports-league management, ride sharing, data clustering, and in assigning students to group projects. In this paper, we study a mathematical-programming approach to "team formation" focused on the interplay between two of the most common objectives considered in the related literature: economic efficiency (i.e., the maximization of social welfare) and game-theoretic stability (e.g., finding a core solution when one exists). With a weighted objective across these two goals, the problem is modeled as a bi-level binary optimization problem, and transformed into a single-level, exponentially sized binary integer program. We then devise a branch-cut-and-price algorithms and demonstrate its efficacy through an extensive set of simulations, with favorable comparisons to other algorithms from the literature.