The Network Game: Analyzing Network-Formation and Interaction Strategies in Tandem
Proceedings of the IEEE/WIC/ACM International Conference on Intelligent Agent Technology, pages 184-192, December 2008.
In networks, autonomous nodes have action choices along two dimensions: They can forward/process incoming service requests – or not, and they can establish new links and maintain or terminate existing ones. In other words, a node can choose both an action-selection and a linkselection strategy. Nodes in networks are either humans (as in social networks) or agents controlled by humans, so the question how humans forming networks behave is important. Humans behave boundedly rational at times, e.g., game-theoretic predictions do not always hold. This paper investigates the questions whether individuals being nodes in a network find and actually use strategies leading to efficiency (in the economic sense), how exogenous parameters influence behavior, and what are characteristics of the resulting network. We do so by carrying out laboratory experiments with human subjects (120 altogether) who form a network and have those action choices. Our work is different from previous work on P2P systems which lets nodes choose their actions or their links, but not both. Main findings are that link selection increases cooperation and results in optimal payoffs in relative terms. Thus, link selection fosters efficiency.