Abstract
The synthetic environment for analysis and simulations (SEAS) is a computational experimentation environment that mimics real life economies, with multiple interlinked markets, multiple goods and services, multiple firms and channels and multiple consumers, all built from the ground up. It is populated with human agents who make strategically complex decisions and artificial agents who make simple but detail intensive decisions. These agents can be calibrated with real data and allowed to make the same decisions in this synthetic economy as their real life counterparts. The resulting outcomes can be surprisingly accurate. This paper discusses the research in this area and goes on to detail the architecture of SEAS. It also presents a detailed case study of market and supply-chain co-design for business-to-business e-commerce in the PC industry.
Similar content being viewed by others
Notes
Details of the exact functional forms are available upon request. What follows is a highly condensed description.
References
Bailey J, Bakos Y (1997) An exploratory study of the emerging role of electronic intermediaries. Int J Electron Commerce 1(3):7–20
Bhargava HK, Vidyanand C (2004) Economics of an information intermediary with aggregation benefits. Inf Syst Res 15(1):22–37
Bichler M, Segev A, Zhao JL (1998) Component-based e-commerce: assessment of current practices and future directions. ACM Sigmod Record: Special Section on Electronic Commerce 27(4):7–14
Carriero N, Gelertner D (1989) Linda in context. Commun ACM 32(4):444–458
Chaturvedi AR, Gulati RK (1993) Computational ecology of manufacturing systems, workshop on information technologies and systems. Orlando, Florida
Chaturvedi AR, Mehta S (1999) Simulations in economics and management: using the SEAS simulation environment. Commun ACM 42(3):60–61
Chaturvedi AR, Nazareth D (1998) Adaptive configuration management for manufacturing systems control. Int J Intell Syst Account Finance Manag (page references needed. This would have appeared.)
Chaturvedi AR, Hutchinson GK, Nazareth D (1993) Supporting complex real-time decision making through machine learning. Decision Support Syst 9:1–21
Chaturvedi AR, Mehta SR, Dolk D, Ayer R (2004) Artificial labor market. Eur J Operat Res (forthcoming)
Chavez A, Maes P (1996) Kasbah: an agent marketplace for buying and selling goods. In: Proceedings of the 1st international conference on the practical application of intelligent agents and multi-agent technology (PAAM’96), London
Downs-Martin J (1997) Final report on the progress of the behavioral taxonomy project (BTP) sponsored by the defense modeling and simulation office (DMSO). Institute for Defense Analyses working paper. IDA, Alexandria, p 3155
Duffy J (2001) Learning to speculate: experiments with artificial and real agents. J Econ Dyn Control 25(3–4):295–319
Epstein JM, Axtell R (1996) Growing artificial societies; social science from the bottom up. The Brookings Institution and the MIT Press, Washington
Fensel D, Ding Y, Omelayenko B, Schulten E, Botquin G, Brown M, Flett A (2001) Product data integration for B2B e-commerce. IEEE Intell Syst 16(4):54–59
Fudenberg D, Levine DK (1999) The theory of learning in games. MIT Press, Cambridge
Gode DK, Sunder S (1993) Allocative efficiency of markets with zero intelligence (ZI) traders: market as a partial substitute for individual rationality. J Polit Econ 101:119–137
Gode DK, Sunder S (1997) What makes markets allocationally efficient? Q J Econ 112:603–630
Green K (2002) Forecasting decisions in conflict situations: a comparison of game theory, role-playing, and unaided judgement. Int J Forecast 18(3):321–344
Guttman RH, Moukas AG, Maes P (1998) Agent-mediated electronic commerce: a survey. Knowl Eng Rev 13(2):147–159
Hedberg (1996) Agents for sale: first wave of intelligent agents go commercial. IEEE Expert 11(6):16–19
Kagel JH, Roth A (1995) Handbook of experimental economics. Princeton University Press, New Jersey
Kaplan S, Sawhney M (2000) E-hubs: the new B2B marketplaces. Harv Bus Rev 78(3):97–103
Kreps DM (1990) A course in microeconomic theory. Princeton University Press, Princeton
Lopez-Paredes A, Hernandez-Iglesias C, Gutierrez JP (2002) Towards a new experimental socio-economics: complex behaviour in bargaining. J Socio Econ 31(4):423–429
Lucking-Reiley D, Spulber DF (2001) Business-to-business electronic commerce. J Econ Perspect 15(1):55–68
MasCollell A, Whinston AD, Green JR (1995) Microeconomic theory. Oxford University Press, New York
Maturana FP, Norrie DH (1996) Multi-agent mediator architecture for distributed manufacturing. J Intell Manufact 7:257–270
Mullainathanan S, Thaler R (2000) Behavioral economics. NBER working paper 7948
Murphy-Hoye (2002) Optimizing the extended enterprise in the new economy—B2B in supply chain management. In: Paper presented at the Center for E-Business at MIT, Annual conference, April 17–19
Nagurney A (2002) Supply chain networks and electronic commerce: a theoretical perspective. Netnomics 10(2):187–220
Plott CR, Gray P (1990) Multiple unit double auction. J Econ Behav Organ 13(2):245–258
Rust J (1996) Dealing with the complexity of economic calculations. Yale University, Mimeo
Sen R, King RC, Xia M (2004) Impact on web-based e-commerce on channel strategy. Int J Electron Commerce (forthcoming)
Shubik M (1975) Games for society, business and war. Elsevier, Amsterdam
Swenson KD, Maxwell RJ, Matsumoto T, Saghari B, Irwin K, (1998) A business process environment supporting collaborative planning. J Collaborative Comput 1(1)
Timmers P (1999) Electronic commerce: strategies and models for B2B trading. Wiley, New York
Tomak K, Xia M (2002) Evolution of the B2B marketplaces. Electronic Markets 12:84–91
Tomak K, Xia M (2003) A strategic analysis of exchange based B2B networks. In: Shaw M (ed) Kluwer, Boston
Xia M, Koehler GJ, Whinston AB (2004) Pricing combinatorial auctions. Eur J Operat Res 154(1):251–271
Zhao K, Shaw M, Xia M (2004) The structure evolution of B2B enterprise networks. In: Tomak K (ed) Advances in the economics of information systems. Idea Group Publishers, Hershey
Acknowledgements
This research was funded by National Science Foundation grants # EIA-0075506 and DMI-0122214. As required by the Memorandum of Understanding between the authors and Purdue University, it is disclosed that some or all of the intellectual property described herein may be commercialized.
Author information
Authors and Affiliations
Corresponding author
Appendix
Appendix
1.1 SEAS virtual execution environment
An Internet-based SEAS virtual execution environment (VEE) is depicted in Fig. 8. In SEAS’ VEE, participants from anywhere can take part in an exercise. The VEE consists of three classes of servers: application servers, distributed database servers, and proxy servers. Proxy servers ensure restricted access for the subscribers. There are three different classes of application servers. The agent-processing servers are capable of running hundreds of thousands of different kinds of agents. Economic processing servers are capable of representing different types of markets. Finally, the visualization servers generate advanced 3-D displays of the data generated during the exercise.
There are individual database servers that support each of these application servers. These database servers may run at one or more locations. Thus, the distributed design enables any numbers of participants to take part in an exercise, and many exercises may be available at any given time.
The SEAS software environment is highly reconfigurable. It consists of three classes of active objects called “bots” as depicted in Fig. 9. The ServBots are autonomous agents that perform the business tasks. Examples of ServBots are BuyBot (agents that perform buying function), SellBot, AcquireBot, ProduceBot, etc. These bots also function as application programming interface (API) for third party development.
The second class of bots is SysBots. This class of autonomous agents performs system level tasks such as read, write, update, open, and close connections.
The third class of bots is the data bots. These bots are being developed to interface with enterprise systems so that firms can seamlessly integrate with SEAS to run dynamic exercise to explore new strategy spaces or may use it as a wind tunnel of corporate strategies. These bots interact with the data manager for their data needs using the SEAS middleware.
The VEE interacts with reconfigurable interface classes. There are eight types of interface classes—ticker, display, system control, local control, global control, user, role, and feedback. ServBots, SysBots, and DataBots are dynamically assembled for each of these interface classes based on the participants profiles or demands by an application synthesizer. This architecture provides the necessary flexibility to adapt SEAS to a wide variety of problem domains.
1.2 SEAS agent architecture
SEAS use intelligent agents (IA) to represent economic realities of electronic markets and hierarchies in a decentralized manner. Intelligent agents in SEAS are autonomous processes that are adaptive and behave like human agents in a narrow domain. In their respective domains, each agent has a well-defined set of responsibilities and authorities so that it can execute its tasks effectively. Examples of SEAS’ IAs are: economic agents—consumer IAs, producer IAs, and regulator IAs and political agents—government IAs, special interest IAs, etc. An agent in SEAS is equipped with reasoning, action, and communication skills required for performing their respective tasks. A SEAS IA is characterized by the knowledge it possesses.
Each agent has a set of seven behavior primitives that enable him or her to perform actions autonomously as shown in Fig. 10. These primitive SysBots are used to initiate, search, decide, execute, update, communicate, and to terminate. Different algorithms for SysBots are used to differentiate between the agents. For example, an agent representing a smart individual will have a more sophisticated search and evaluation algorithm than that of a not so smart agent. There can be several instantiations of each SysBot. SysBots that constitute an agent are described below:
-
Initiate: these bots are the sensors for a given ServBot. When certain conditions are met then they trigger action from the ServBot.
-
Search: these bots provide the ability to search the space for best price, quantity, etc.
-
Evaluate: these bots enable the servbot to evaluate different alternatives and select the most appropriate course of action.
-
Execute: these bots execute the course of actions needed by the ServBot.
-
Communicate: these bots have the knowledge of the workflow and the chain of command. After the action is taken, these bots communicate the appropriate message/information to the appropriate parties.
-
Update: these bots update the relevant information/data at the appropriate times. These bots are critical for the system performance.
-
Terminate: These are quality assurance bots that make sure that the tasks are completed satisfactorily.
1.3 SEAS market architecture
SEAS markets are implemented in JavaSpace. Javaspace is a descendant of Linda system developed at Yale University (Carriero and Gelertner 1989). Java Space defines the market structure as shown in Fig. 11. The four different types of markets structures supported by SEAS are: posted price, double auction, single auction, and reverse auction. The rules for all the markets are implemented through JavaSpace, which in turn synchronizes the thread between the agents. Agents maintained in the space are updated through their working status. JavaSpace also forms the connectivity between the Enterprise Java Beans (EJB) and the Database, and therefore, after completing the transactions, it updates the database.
JavaSpace supports simultaneous running of multiple games. Every agent created in the EJB has an AgentSpace object. AgentSpace gives every agent partial access to the JavaSpace. This way the agent does not have to worry about the implementation of the space. It gives a clean abstract and encapsulated cover to the implementation of the space.
Rights and permissions
About this article
Cite this article
Chaturvedi, A.R., Mehta, S.R., Dolk, D. et al. Computational experimentations in market and supply-chain co-design: a mixed agent approach. ISeB 4, 25–48 (2006). https://doi.org/10.1007/s10257-005-0021-6
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10257-005-0021-6