Acquisition of reciprocal altruism in a multiagent system

This article shows a method to acquire reciprocal altruism. At first, we define a task assignment problem and the problem is modeled as a multiagent system. Then, we propose a method to acquire rules for agents´ decision making. Our method is based on Q-learning. To acquire reciprocal altruism, we use two kinds of feelings, one is to be grad of other agents´ happiness and the other is to sympathize with unhappiness of other agents. These feelings are defined as positive and negative rewards when agents learn rules for their decision making. The proposed method has been implemented and some experiments have been done. Experimental results show that an appropriate negative reward as pity contributes to acquisition of reciprocal altruism.