A Mechanism of Dynamical Interactions for Two-Person Social Dilemmas

We propose a new mechanism of interactions between game - theoretical
agents in which the weights of the connections between interacting
individuals are dynamical, payoff - dependent variables.
Their evolution depends on the difference between the payoff of the agents
from a given type of encounter and their average payoff. The mechanism is
studied in the framework of two models: agents distributed on a random
graph
and a mean field model. Symmetric and asymmetric connections between the
agents are introduced. Long time behavior of both systems is discussed for
the Prisoner's Dilemma and the Snow Drift games.