On the intrinsic coordinatability of network control systems

Concepts of both group and individually rational behavior are essential for a control network to cooperate in the true sense. Although classical game theory provides a powerful framework within which to study networks, it solution concepts are based on, and only on, individual rationality, and do not admit notions of group-level rational behavior. Conditional game theory overcomes this limitation. In contrast to conventional categorical utilities, which encode agents' preferences as functions of the possible actions of themselves and others, conditional utilities permit agents to condition their preferences on the preferences of other agents as well as their actions. As these conditional utilities propagate through the network, they create a social model which leads to an emergent notion of system level rationality. This expanded formulation of game theory also permits the assessment of the innate ability of a network system to coordinate its behavior. Shannon information theory is adapted to define an explicit measure of the intrinsic capacity for a system to coordinate. This measure may be used to assess the ecological fitness of the agents to function in their environment.