Towards Efficient Equilibria of Combinations of Network-Formation and Interaction Strategies

Abstract

Abstract-Agents in networks have two strategic choices: They can forward/process incoming service requests -or not, and they can establish additional contacts and maintain or terminate existing ones. In other words, an agent can choose both an action-selection and a link-selection strategy. So far, it is unclear which equilibria exist in such settings. We show that there are the following equilibria: First, an inefficient one where agents leave the network. Second, an equilibrium where agents process requests on behalf of others, i.e., they cooperate. In this second equilibrium, agents distribute their contacts uniformly, which is not efficient. We show that a strategy, we propose in this paper, yields an equilibrium that is optimal, i.e., that yields the highest sum of payoffs over all equilibria. If agents base their link-selection decisions on the processing times of their requests, optimal system states can be equilibria.

Strategic Behavior in Interaction Selection and Contact Selection

“…This leads to cooperative networks, even though no reputation system has been used. The result confirms our theoretic findings in [14]. There, we show under which conditions payoff-maximizing players form efficient systems.…”

“…In [14], we have investigated search in coordinator-free networks of payoff-maximizing participants. We now briefly present the most important results again, since they serve as a reference point regarding the strategies we might observe in real systems.…”

“…Our theoretic analysis of interaction and contact selection in [14] shows that -in theory -two players cooperate under the following conditions:…”

“…(a) and (b) grow with the number of hops, and the number of hops depends on the number of additional contacts of a participant and on their distribution. In [14], we have shown that payoffmaximizing players have an incentive to form networks that allow processing requests with few hops only. In this paper, we are interested in the question whether human behavior leads to similar results, i.e., if game theory leads to meaningful predictions in interaction and contact selection.…”