What

This project aims to study various ways of designing intelligent trading and market-making agents. We are simultaneously devising and refining an artificial market environment, and using it to study cooperative and competitive agent behavior.

Why

In the last decade there has been a surge of interest within the finance community in describing equity markets through computational agent models [7]. At the same time, financial markets are an important application area within artificial intelligence for the fields of agent-based modeling and machine learning, since agent objectives and interactions tend to be more clearly defined, both practically and mathematically, in these markets than in other areas. Computational modeling of markets allows for the opportunity to push beyond the restrictions of traditional theoretical models of markets through the use of computational power while also allowing a fine-grained level of experimental control that is not available in real markets. Thus, data obtained from artificial market experiments can be compared to the predictions of theoretical models and to data from real-world markets, and the level of control allows one to examine precisely which settings and conditions lead to the deviations from theoretical predictions usually seen in the behavior of real markets. This project is also an application area for the general problems of distributed intelligence such as collective learning, coordination and competition. We are interested in studying how software agents endowed with learning abilities might interact, co-evolve, and cooperate in societies of learning agents.

Previous Work

The project draws on at least three distinct subfields: market microstructure, experimental markets, and simulated markets. Studies in market microstructure theory provide important background and context for the experiments and simulations [8]. An alternative to the theoretical approach is an experimental one in which individuals are placed in a controlled market setting, given certain endowments of securities and cash, and allowed to trade with each other [1,5]. Lastly, computer simulations of markets populated by software agents extend the experimental approach by allowing the experimenter to test various theories of learning behavior and market microstructure in a controlled environment([4,2,7] inter alia).

How

Within the artificial markets project, we are pursuing four major research directions, some of which are discussed in detail in companion abstracts.

1. Questions about financial markets:
   • What is the role of bounded rationality (agents with limited capabilities)?
   • How do bounded rationality markets compare with the perfect information, fully rational agents of classical equilibrium theories?
   • What is the role of market rules and mechanisms (for example clearing, specialists, continuous double auctions vs. sealed double auction markets)?
   
   
2. Questions about market dynamics and the evolutionary environment:
   • What are the properties of markets populated with traders that use many different types of utility functions? Which of these utility functions are evolutionarily successful and stable?
   • What is the role of selection by competition with other agents and from the environment?
   • What is the role of various mutation mechanisms?
   
   
3. Questions about cognitive behavior:
   • What is the effect of certain cognitive limitations on market behavior (the direct problem)?
   • What can we infer from the overall market about the cognitive abilities of individual agents (the inverse problem)?
   • What is the simplest realistic model of a financial agent?
   
   
4. Questions about learning agents:
   • How do we develop and compare specific learning algorithms? What should the properties of these algorithms be?
   • Can seemingly collusive behavior in the financial market setting emerge as a natural result of learning by traders and dealers?

Also, we are continuing to develop and improve our artificial market software, which provides a testbed for many experiments and allows artificial and human agents to interact with each other in a controlled setting.

Impact

This research will provide insights into the impact of market structure and design on market efficiency, patterns of information flow and price processes. For example, we hope to investigate the advantages of monopolistic vs. competitive dealer markets given different assumptions about market volatility. Simultaneously, investigating automated trading algorithms will provide insight into the kinds of environments where particular trading strategies may be successful. The questions we ask in this project lie at the intersection of several disciplines, from computer science (distributed systems of agents), to learning (which is a key aspect of the artificial agents and possibly also of the market structure), to economics (financial markets are the primary focus), to cognitive sciences (interaction between agents' biases and properties with the overall behavior of the market).

Future Work

We are researching more sophisticated learning algorithms for our agents and the dynamics created by heterogeneous preferences. In the long term we are looking into the stability and evolutionary dynamics of different learning strategies in societies of agents. Additionally, we will study the possible refinement of learning techniques to deal with more complex market environments.

Research Support

This research was sponsored by grants from: Office of Naval Research (DARPA) Contract No. MDA972-04-1-0037, Office of Naval Research (DARPA) Contract No. N00014-02-1-0915, National Science Foundation (ITR/IM) Contract No. IIS-0085836, National Science Foundation (ITR/SYS) Contract No. IIS-0112991, National Science Foundation (ITR) Contract No. IIS-0209289, National Science Foundation-NIH (CRCNS) Contract No. EIA-0218693, National Science Foundation-NIH (CRCNS) Contract No. EIA-0218506, and National Institutes of Health (Conte) Contract No. 1 P20 MH66239-01A1.

Additional support was provided by: Central Research Institute of Electric Power Industry, Center for e-Business (MIT), Daimler-Chrysler AG, Compaq/Digital Equipment Corporation, Eastman Kodak Company, Honda R&D Co., Ltd., ITRI, Komatsu Ltd., Eugene McDermott Foundation, Merrill-Lynch, Mitsubishi Corporation, NEC Fund, Nippon Telegraph & Telephone, Oxygen, Siemens Corporate Research, Inc., Sony MOU, Sumitomo Metal Industries, Toyota Motor Corporation, and WatchVision Co., Ltd..

References

[1] D. Davis and C. Holt. Experimental Economics. Princeton University Press, Princeton, NJ, 1993

[2] D. Friedman and J. Rust. The Double Auction Market Institutions, Theories, and Evidence. Technical Report, Santa Fe Institute Studies in the Sciences of Complexity, 1991.

[3] L.R. Glosten and P.R. Milgrom. Bid, Ask and Transaction Prices in a Specialist Market with Heterogeneously Informed Traders. In Journal of Financial Economics, 14:71--100, 1985.

[4] D. K. Gode and S. Sunder. Allocative Efficiency of Markets with Zero Intelligence Traders: Markets as a Partial Substitute for Individual Rationality. In Journal of Political Economy, pp. 119--137, 1993.

[5] J. Kagel and A. Roth. Handbook of Experimental Economics. Princeton University Press, Princeton, NJ, 1995.

[6] Albert S. Kyle. Continuous Auctions and Insider Trading. In Econometrica, 53(6):1315--1336, 1985.

[7] B. LeBaron. Agent-based Computational Finance: Suggested Readings and Early Research. In Journal of Economic Dynamics and Control, 24:679--702, 2000.

[8] A. Madhavan. Market Microstructure: A Survey. In Journal of Financial Markets, pp 205--258, 2000.