Game-theoretic cooperativity in networks of self-interested units

The behavior of theoretical neural networks is often described in terms of competition and cooperation. I present an approach to network learning that is related to game and team problems in which competition and cooperation have more technical meanings. I briefly describe the application of stochastic learning automata to game and team problems and then present an adaptive element that is a synthesis of aspects of stochastic learning automata and typical neuron‐like adaptive elements. These elements act as self‐interested agents that work toward improving their performance with respect to their individual preference orderings. Networks of these elements can solve a variety of team decision problems, some of which take the form of layered networks in which the ‘‘hidden units’’ become appropriate functional components as they attempt to improve their own payoffs.