Evolving Specialization, Market and Productivity in an Agent-Based Cooperation Model Erbo Zhao, Guo Liu, Dan Luo, Xing’ang Xia, and Zhangang Han ⋆ Department of Systems Science, School of Management, Beijing Normal University, Beijing, 100875, People’s Republic of China zhan@bnu.edu.cn Abstract. This paper introduces an agent-based model in which self- interest intelligent agents are adaptive. Agents can either go to find re- sources in the environment or mine the resources found. Agents trade information about resources in a market. A biased learning mechanism is introduced to update agents’ capabilities of mining and searching. The learning mechanism plays a vital role in the specialization process in our model. Expectation is also introduced in this paper to determine the trade price. Simulations show that agents can specialize in available ca- pabilities, form market and cooperate to increase their wealth. These emergencies come out through just pre-defining some learning and pric- ing mechanisms that are not so complex but close to reality. Total pro- ductivity and market formation are tracked during the evolving process. The wealth distribution during whole evolving process also demonstrates an interesting power law distribution. Keywords: Agent-based model, individual reinforcement learning, power law, specialization, market forming, expected productivity. 1 Introduction Agent-based models are increasingly recognized as powerful tools for modeling complex adaptive systems. They are useful in the terms of modeling of the inter- actions among agents, representation of learning in a dynamic strategy context, and description of the relationships between individual and its local environ- ment [14]. Agent-based models can track individual interactions that otherwise cannot be represented in traditional models that use global variables. In partic- ular, agent-based models excel at relating the heterogeneous behavior of agents with different information, different decision rules, and different situation to the macro behavior of the overall systems [15]. Recently, agent-based models have been applied to a variety of systems, such as biological ecosystems, economics systems [1, 3, 9, 13, 17, 18] and social intelligence systems [4]. ⋆ Correspondence author. Tel. 86.10.58807876. This research is supported by National Science Foundation of China under grant No. 60774085 and No. 70601002. J. Zhou (Ed.): Complex 2009, Part II, LNICST 5, pp. 1564–1574, 2009. c  ICST Institute for Computer Sciences, Social Informatics and Telecommunications Engineering 2009