Incentive Control in Network Anti-Coordination Games with Binary Types
- Additional Document Info
- View All
© 2018 IEEE. We investigate control strategies that maximize disagreement in type-based anti-coordination network games. Each node is a player and each edge represents payoff dependence between the neighboring players where, depending on their types, have an incentive to play different strategies. We assume there is a preferred strategy in the absence of network effects. Players follow a distributed learning dynamics based on the process of iterated elimination of dominated strategies. In this setting, we propose a disagreement maximization problem. First, we seek to find the minimum number of players to control while ensuring the dynamics converge to maximum possible disagreement. We characterize the control strategies in line networks that achieve optimal results, and propose an algorithm for general networks that are approximately optimal.
author list (cited authors)
Paarporn, K., & Eksin, C.