|
Pang Wei; Ph. D Proposal; Dept. of Computer Science,University of Maryland, College Park.
(I) Proposed Title
(II) Introduction
(III) Brief Literature Review
(IV) Methodology
(V) Proposed Research Time-Table
(VI) References
(I) Proposed Title
Agent Based Simulation and Its Application in Computational Economics
(II) Introduction
Complex Systems
The definition of complex systems provides a new unity of approach to many different problems. These concepts originate from efforts to understand physical, biological and social systems. Examples of applications can be found in all fields and professions including science, medicine, engineering, management and education. It should be pointed out that complex systems is an active field and the most exciting discoveries are yet to be made. Numerous complex systems are vital to human beings such as market economics, social systems, immune system and etc. So it is important to understand and analyze them well.
The Agent System and Agent Based Simulation
The general evolving agent system is made up of many autonomous agents and each agent can interact with others. Given certain initial conditions, the agent system begins to evolve without external intervention. After several iterations, some meaningful results could be found in this closed system. So the agent based simulation is utilized to investigate the common encountered complex systems especially the social and economical systems.
Agent based Computational Economics
ACE (Agent-based Computational Economics) is a brand new cross research area of economics and computational intelligence. It is the computational study of economies modeled as evolving systems of autonomous interacting agents. Decentralized market economies are complex systems, and the agents might represent people, companies interacting in market environment composed of resources, trading protocols, technology level, etc.
Economics is a well developed discipline; economists employed a lot of methods to study this field. In the past 10 years, with the quick development of computing science and technology, some economists used a new technique—constructed method from bottom to up implemented by agents simulating economic entries' action in markets. This gave us a new way to explore the distributed market economies and make made an in-depth study of the other complex systems. More detailed description about computational economics can be seen in the following website: wuecon.wustl.edu/sce/
(III) Brief Literature Review
Robert Axelrod's work
The work of Axelrod can be considered as the preliminary research on in the area of agent based social simulation.
In Axelrod's book the Evolution of Cooperation [1], the author discussed how cooperation can emerge in a world of self-seeking egoists--whether superpowers, businesses, or individuals--when there is no central authority to police their actions. The latter book [2] continues the study and pointed out the complexity of the cooperation between interacting agents through gathering together the myriad fruits of a decade's previous work and carefully modified his initial model in the first book.
One of the most exciting aspects is that the author tried to apply the idea of evolutionary computation to the agent based simulation. The simulation environment was looked viewed as an evolving system. The most common used evolutionary algorithm-Genetic Algorithm was employed in the environment to investigate the old classic problem: iterated prisoner's dilemma problem.
This inspires us that the computational intelligence theory can be applied to the evolving agent system and it is helpful to construct the simulation environment and design the culture dish experiment successfully.
Leigh Tesfatsion's work
The work of Leigh Tesfatsion made a major contribution to agent based computational economics by developing the culture dish approach to construct an agent system and to explore the hidden rules among the interacting agents. She summarized the main research areas in ACE: (i) Learning and the embodied mind; (ii) evolution of behavioral norms; (iii) bottom-up modeling of market processes; (iv) formation of economic networks; (v) modeling of organizations; (vi) design of computational agents for automated markets; (vii) parallel experiments with real and computational agents; and (viii) building ACE computational laboratories.
Tesfatsion illustrated that current ACE research divides roughly into four strands differentiated by objective:
1.One primary objective is empirical understanding: Why have particular macro regularities evolved and persisted, despite the absence of top-down planning and control?
2. A second primary objective is normative understanding: How can agent-based models be used as laboratories for the discovery of good economic designs?
3.A third primary objective is qualitative insight and theory generation: How can the full potentiality of economic systems be better understood through a better understanding of their complete phase portraits (equilibria plus basins of attraction)?
4.A fourth primary objective is methodological advancement: How best to provide ACE researchers with the methods and tools they need to undertake the rigorous study of economic systems through controlled computational experiments?
The study of Complex Network and Social Network
Recently, the theory of complex network [6] is developed quickly because of the emergence of many new type complex network systems and the recognition of numerous classic complex systems which can be seen as a complex network. Complex weblike structures describe a wide variety of complex systems of high technological and intellectual importance such as Internet.
The human social network can be seen as a complex network on which fads and ideas spread. The nodes of the social network are human beings and edges represent various social relationships. Social network analysis (SNA) [7] is focused on uncovering the patterning of people's interaction.
Motivated by the above research and recognition, some quantities and measures which embody the complex network theory and SNA may be proposed and investigate in depth in the field of economics together with the agent based simulation.
(IV) Methodology
Model Construction
Simulation Approach
Our research group is working on an ACE model now. Our main goal is to research behavior of exchange between agents with various product ability and different utility functions. This model is inspired by reference [4], [5], we just considered commodity exchange in a simple simulation society. Commodity exchange is a basic economics behavior, and is the origin of market economies, so it is valuable for economics research. Only considering commodity exchange is important to make our model not too complex to be implemented. We hope some valuable patterns could be found in our simulation by this model. The focus of the research is how exchange improves overall welfare, how exchange cost affects exchange, etc.
Basing on this simple model, more complex model could be built by carefully inserting additional constraints and conditions. Once a specified model has been constructed, specified phenomenon could be observed and interesting patterns could be found.
We intend to use the Darwin's evolutionary theory and natural selection mechanisms to make the agents cope with the artificial simulation system. The detailed design for the simulation model may be done in the future work.
Complex Network Analysis Approach
As we have mentioned before, the analysis method of complex network could be applied to the agent based simulation environment. Each agent can represent a node in the complex network, what will happen if the agents possess the characters properties of the node in the network? On the other hand, whether the agent based simulation environment can be seen as a complex network if we add some rules and constraints? Is the virtual market economics environment also a kind of complex network such as small world network, evolving network or even scale free network [6] if specified conditions are satisfied? The above considerations are significant questions and valuable to study and verify.
Data Collection and Analysis
Currently the initial model has been constructed and some rough data can be reported. In the future work, the data format should be standardized and some visualized results will be presented before the modelers. With these useful data collected from the framework which is based on the model we have proposed, some algorithms and analysis methods can be utilized to find significant rules. For instance, the whole evolving procedure of the agent system can be recorded to a file and we can clearly see the variation of the agent and the tendency of the overall system at a specified time or iteration.
Verification and Conclusion
Once we had drawn some conclusions which we considered as an important theory, we have to testify our theory and hypothesis by some certain methods. The methods could include both the established theory proven one or the real experiment one. The newly developed Experimental Economics may also provide us an efficient solution, but the concrete deployment steps will be a challenge question and much work should be done.
On the other hand, the theory which was tested to be validate can guide the real market economics and other aspects in the economics, thus to make us better understand the hidden rules in the macro economical environment. This is the ultimate goal of the agent based simulation on Computational Economics.
(V) Proposed Research Time-Table
Sept 2005----Apr 2006 : - Literature review, familiar with the related research area, Obtain a comprehensive knowledge on Agent based social simulation.
May 2006----Jun 2007 -Field Work- model Design: First propose some hypothesis model, and then discuss the validity of the model in the theory aspect. Design the necessary algorithms, analysis the complexity of the model.
Jul 2007----Sept 2007: - Field Work - Construct the model. While studying the proposed models, construct some potential excellent models, coding and testing.
Oct 2007---Nov 2007: - Field Work- Data Analysis: Modify and perfect the model, in the same time collect the data from the already build model, acquire useful information, and then preprocess, analysis and visualize the data by certain techniques thus to find useful patterns and rules.
Dec 2007----Feb 2008: - Field Work – Model Verification: testify the model by theory, real world or experiment. Summarize the results.
Mar 2008----Aug 2008 : - Write the Thesis
(VI) References
[1]Agent-Based Computational Economics, LEIGH TESFATSION, http://www.econ.iastate.edu/tesfatsi/ ,ISU Economics Working Paper No. 1, Revised August 24, 2003
[2] Evolution of Cooperation, Robert M. Axelrod, Princeton University Press, Princeton, New Jersey, ISBN: 0465021220
[3] The complexity of cooperation-Agent Based Models of Competition and Collaboration, Princeton University Press, Princeton, New Jersey, ISBN: 0691015678
[4] Learning to speculate: Experiments with artificial and real agents, John Duffy,Journal of Economic Dynamics and Control ,Volume 25, Issues 3-4 , March 2001, Pages 295-319
[5]On Money as a Medium of Exchange,Kiyotaki, Nobuhiro Wright, Randall,Journal of Political Economy,Volume (Year): 97 (1989),Pages: 927-954
[6] Statistical mechanics of complex networks, Reka Albert, Albert-Laszlo Barabasi,Reviews of Modern Physics 74, 47 (2002)
[7] Introduction to Social Network Methods, Robert A.Hanneman, faculty.ucr.edu/~hanneman/SOC157/NETTEXT.PDF
|
