ICES: Small: Contribution Games in Social Networks

Rensselaer Polytechnic Institute, Troy NY

Investigators

Abstract

This project focuses on network formation by self-interested agents. Specifically, it studies contribution games in social networks, where players/nodes not only form connections/relationships with others, but also determine the intensity of their connections. This project will analyze the quality of equilibrium solutions, convergence properties of local dynamics, and mechanisms to improve these properties. This work will result in efficient algorithms for computing various reasonable outcomes of contribution games, and algorithms for "nudging" the players to form better outcomes by giving them small incentives, using approximation algorithms for otherwise intractable problems. More concretely, this research will consider the following topics. (1) Analyzing a model of Network Contribution Games that captures the essence of players optimizing their local relationships. Concepts such as Nash equilibrium are of limited usefulness for such models, so the project instead focuses on pairwise equilibrium, strong equilibrium, and convergence dynamics. (2) General Contribution Games are an important extension of network contribution games: instead of pairwise relationships, players contribute to many-person projects. (3) Centrality, Friendship, and Altruism. Other than the desire to maximize total success of all projects/relationships players are involved in, they may also desire to form connections that make them well-positioned in the social network, and so attempt to maximize some notion of "centrality" or they may care about the quality of the network as a whole, as in games with public goods. This project will analyze games where the utility of a player is a combination of such considerations, and derive general results on the structure and quality of equilibrium solutions depending on the types of player utility functions. Social networks pervade most aspects of human life, and affect everything from the spread of epidemics to the ability to find a job. Now that data on social networks has become easier to collect, understanding the structure, properties, and incentives of social network formation has become a crucial research topic. By explicitly taking into account the strength of bonds in a social network, and not just their existence, this work will add a new dimension to game-theoretic models of social networks, and obtain new insight into participant behavior, and into the structure of social networks. This work may also suggest efficient methods for "nudging" the participants of a social network to form better outcomes by giving them small incentives. In addition, this project will contribute new techniques for analyzing systems with limited collusion. The study of contribution games where participants care about centrality, friendship, and altruism in addition to their local success, will be the first major attempt to understand the structure of social networks under such a combination of player incentives. This project will include collaborations with both theoretical and applied researchers working on social networks, including researchers from complex systems, epidemiology, sociology, and the school of management. Because of wide-spread interest in social networks, this project may inspire new approaches in many of the fields mentioned above. Finally, this research will be strongly complimented by the PI's education plan, which includes teaching several courses with research components, forming a workshop on this topic, and recruiting several graduate and undergraduate students to work on this project.