Social structure optimization in team formation.

This paper presents a mathematical framework for treating the Team Formation Problem explicitly incorporating Social Structure (TFP-SS), the formulation of which relies on modern social network analysis theories and metrics. While recent research qualitatively establishes the dependence of team performance on team social structure, the presented framework introduces models that quantitatively exploit such dependence. Given a pool of individuals, the TFP-SS objective is to assign them to teams so as to achieve an optimal structure of individual attributes and social relations within the teams. The paper explores TFP-SS instances with measures based on such network structures as edges, full dyads, triplets, k-stars, etc., in undirected and directed networks. For an NP-Hard instance of TFP-SS, an integer program is presented, followed by a powerful Lin-Kernighan-TFP (LK-TFP) heuristic that performs variable-depth neighborhood search. The idea of such λ-opt sequential search was first employed by Lin and Kernighan, and refined by Helsgaun, for successfully treating large Traveling Salesman Problem instances but has seen limited use in other applications. This paper describes LK-TFP as a tree search procedure and discusses the reasons of its effectiveness. Computational results for small, medium and large TFP-SS instances are reported using LK-TFP and compared with those of an exact algorithm (CPLEX) and a Standard Genetic Algorithm (SGA). Finally, the insights generated by the presented framework and directions for future research are discussed. HighlightsWe propose a methodological toolbox that incorporates social structures into TFP.The use of optimization techniques in the area of social science is motivated.The relation between social science findings and team outcomes is explored.Explicit, rigorous functions of social structures to evaluate outcomes are presented.LK-TFP heuristic based on variable depth neighborhood search is proposed for TFP.