This paper models the organization of the firm as a type of artificial neural network in a duopoly setting. The firm plays a repeated Prisoner’s Dilemma type game, and must also learn to map environmental signals to demand parameters and to its rival’s willingness to cooperate. We study the prospects for cooperation given the need for the firm to learn the environment and its rival’s output. We show how profit and cooperation rates are affected by the sizes of both firms, their willingness to cooperate, and by environmental complexity. In addition, we investigate equilibrium firm size and cooperation rates.