International Journal of Academic Research in Economics and Management Sciences May 2013, Vol. 2, No. 3 ISSN: 2226-3624

Understanding Enterprise Behavior Using Simulation Models Ph.D. Student Ramona-Mihaela MATEI Bucharest University of Economic Studies, Romania [email protected]

Ph.D. Professor Ioan RADU Bucharest University of Economic Studies, Romania [email protected] Abstract The high complexity of the systems and the dynamics of the organizational environment generated the interactions of its components requires of managers adaptable, expandable and in real time modeling capabilities of the real state of the company in order to capture the way in which the external and internal factors can affect enterprise behavior. Practice demonstrates that traditional methods cannot provide this capability for quantitative analysis of behavior in a holistic perspective. In this context, there have been developed the tools and simulation models to provide the capability to create a virtual model that can be easily manipulated in hypothesis testing related to the enterprise behavior and the effective assessment of the alternatives and possible scenarios. This paper aims to provide an analysis of the Agent-Based Modeling and Simulation (ABMS) and System Dynamics (SD) in order to contribute to a better understanding of the enterprise behavior and to guide selecting appropriate simulation approach for obtaining a correct analytical result. Keywords: Enterprise Behavior, Simulation Model, Agent-Based Modeling and Simulation, System Dynamics. Introduction Organizations as very complex systems require a greater capacity to adapt in order to survive in a competitive environment. This need requires a high level of understanding and innovation of the problem complexity. The high complexity of the systems and the dynamics of the organizational environment generated the interactions of its components requires of managers adaptable, expandable and in real time modeling capabilities of the real state of the company in order to capture the way in which the external and internal factors can affect enterprise behavior. Practice demonstrates that traditional methods cannot provide this capability for quantitative analysis of behavior in a holistic perspective. In this context, there have been developed the tools and simulation models to provide the capability to create a virtual model that can be easily manipulated in hypothesis testing related to the enterprise behavior and the effective assessment of the alternatives and possible scenarios. Managerial 68

www.hrmars.com

International Journal of Academic Research in Economics and Management Sciences May 2013, Vol. 2, No. 3 ISSN: 2226-3624

simulation of the enterprise behavior is more difficult due to the large complexity of the tasks, the environment uncertainty in which acting and multi-scale nature of it. Due to increased dynamics of an enterprise, using such simulation models require primarily a good knowledge and analysis of this behavior. In this context simulation models allow the systematic investigation and analysis of complex processes dynamic behaviors which cannot be modeled with traditional mathematical methods. The behavior of complex systems is not linear, but rather is the result of feedback and interactions of internal components with external environment factors, which makes behavior modeling to be difficult or impossible to achieve with mathematical approaches based on linear extrapolation of the behavior. In the literature and practice it is recognized that the simulation is a computerized tool able to incorporate the uncertainties inherent in the real complex systems (Keskhin, Melouk, and Meyer, 2010) which leads to operational and managerial efficiency improvement of enterprise decision-makers. There are several types of organizational behavior simulation methodologies in the literature (Cohen and Cyert, 1965; Macy and Willer, 2002 Burton, 2003) that differ in terms of behavioral approaches. The disadvantage of these methodologies consists in difficulty to provide a representative and realistic model compared to a real enterprise behavior. Another category includes most used in practice simulation methodologies as well as models of Discrete Event Simulation, System Dynamics, Agent-Based Modeling and Simulation capable to capture and to simulate a specific dimension of enterprise behavior. In this context, the paper aims to provide an analysis of the ABMS and SD simulation models highlighting the role, the advantages and limitations of using these models. The purpose of this paper is to contribute to a better understanding of the enterprise behavior and to guide selecting appropriate simulation approach for obtaining a correct analytical result. System Dynamics (SD) simulation models Systems Dynamics (SD) is a simulation methodology that was developed by Jay Forrester in 1950 with the aim of macro-scale industrial behavior modeling based on the study of information feedback characteristics of industrial activities and of the demand amplification effects on the supply chain. According to Forrester (1958), SD reveals how the interactions between organizational structure, enhancing policies, actions and decisions delays can influence the enterprise behavior. SD is based on cybernetics and control theory principles and on the use of differential continuous equations combined with charts which gives the capability to identify the causal structure of the system and its influence on behavior. SD is considered a simulation tool capable to perform changes at the highest organizational level and to sustain top management in solving complex problems, especially those characterized by significant delays and feedbacks such as resource allocation and supply chain management problems. Sterman (2000) and Fowler (2003) state that SD method can be used as a tool for organizational learning (Senge, 1990) capable to supports managers in discovering and understanding the specific dynamics of organizational structures. 69

www.hrmars.com

International Journal of Academic Research in Economics and Management Sciences May 2013, Vol. 2, No. 3 ISSN: 2226-3624

SD was used to model the behavior since its first application in modeling of production systems dynamics and industrial supply chain (Forrester 1958). Later, the use of SD was extended in various fields such as: the general theory creation (Repenning, 2002), systems modeling with the goal of forecasting and developing policies (Lyneis, 2000), study of the dynamics strategy (Fowler, 2003), and distribution supply chain simulation (Angerhofer and Angelides 2006, Scheiritz and Grossler, 2003), simulation of product development (Ford and Sterman 1998) or business process, economics, ecology, human resources management, software development, competition, innovation or improving the decision making process of an organization (Miragliotta et al, 2009). SD can models a real system by using basic elements (input and output flows, ”stocks”, variables and feedback loops) by combining of which it is possible to simulate more complex elements such as delays in order to identify how the system can reach the balance state (Lattila et al, 2010). In the SD methodology system behavior is modeled over time through feedback and delays capturing in the process and based on the representation of resources and system dynamics as a set of accumulation ("stocks") and flows between them. In the SD model, the complexity is generated by the interaction of multiple feedback loops and system structure (Maani and Maharaj, 2004). The implementing methodology of SD illustrate the system processes and entities in form of aggregate accumulations/"stocks"(financial resources, materials, knowledge, human resources) represented by integral equations with inflows and outflows, with causal variables that influence such flows and the delays between them. Accumulations are those which drive the system model towards its balance-state. Flows are characterized by feedback and delays so that the resulting performance of the system can be non-linear and sometimes counterintuitive. According to Swinerd and McNaught (2012), flows in the system can meet the average rate of its entities state changes. Under this methodology causal dependencies within the system are represented in the form of diagrams that can be mathematically modeled as a set of differential equations which generate a solving solution for the base system. Agent-based modeling and simulation (ABMS) Agent-based modeling and simulation, “multi-agent systems” or “artificial societies” modeling (Sawyer, 2003) is a relatively new method of simulating dynamic and complex systems distributed in time and space (Lattila et al, 2010). ABMS represents system entities in the form of decentralised, autonomous software agents operating in parallel and communicating between them by means of sets of internal rules in order to produce a pattern of behavior at the system level. This implies that the real system is modeled as a set of agents (agents system) that interact in an environment that is defined and implemented by means of simulation software. A system is comprised of individual agents that have specific relations within a defined environment. In this case complexity is generated by interactions between agents and not by the structure of the entire system (Scholl, 2001). ABMS allows connections capturing from micro to macro level and assess how micro-level interactions may generate behaviors at macro level. 70

www.hrmars.com

International Journal of Academic Research in Economics and Management Sciences May 2013, Vol. 2, No. 3 ISSN: 2226-3624

Axelrod and Tesfatsion (2006) assert that agents-based models are defined on the basis of two criteria: 1) system is composed of interacting agents and 2) emerging properties of the system are generated from the interaction between agents and properties cannot be deduced through the aggregation of individual properties. Agents acting on the basis of an internal schema that contains sets of rules that define the decision-making capabilities of their own, have memory, can teach, it can adapt and can act in the simulation or by changing the status, either by making decisions to act on the basis of inputs from their local environment. Although they have limited capacity to recognize the system as a whole, the agents are able to collaborate and interact between them and with the external environment in order to achieve a certain goal. This system of sharing of information, knowledge and tasks between agents can contribute to the creation of collective intelligence, which would not be possible only on the basis of the internal mechanism of each agent. In the late 1990s, researchers in the social sciences start to become aware of the usefulness of agent-based models for modeling social systems through the generation and validation of theories that can be applied to specific systems (Anderson and Meyer, 1999). For many researchers in this field ABMS are the third way for theory development in addition to the deduction and induction (Axelrod and Tesfatsion, 2006). Agent-based modeling allows building bottom-up models of social phenomena and analysis of these patterns by various rules for the better control of its emergent behavior. Carley (2002) argues that agent-based models are applied in social sciences in two ways: 1) an intelectiv way, in which models are used for testing and generating theories and 2) an emulativ way, through which the models simulate the real organizations with complex dynamics in order to support management efforts. Although the usefulness of agents-based models in simulation of social phenomena is widely recognized in the literature, this methodology can also be applied in other areas. As the SD models, the first application of the ABMS was made with the aim of studying the behavior of complex supply chain (Akkermans, 2001; Scholl, 2001; Schieritz and Grossler, 2003). Subsequently, as a result of the development of these models, the ABMS was used also in the fields of production such as production chain management (Hilletofth and others, 2010), as well as a support in the decision making process (Nillson and Darley, 2006). Garro and Russo (2010) consider the ABMS methodology as an iterative process consisting in seven subsequent phases:  Analysis of the system: this phase involves setting simulation goals and preliminary understanding of the system and its components  Conceptual system modeling: allows the system definition in terms of agents, classes of agents and societies  Simulation design: this phase suppose the construction of system abstraction model as a framework which will be exploited in the simulation  Simulation code generation: the simulation code is generated for proposed simulation based on the model resulted in the previous phase 71

www.hrmars.com

International Journal of Academic Research in Economics and Management Sciences May 2013, Vol. 2, No. 3 ISSN: 2226-3624

  

Establishing scenario simulation Simulation execution Simulation results analysis by comparing them with the objectives established in the initial phase Based on the results obtained in the previous process it can be found a further implementation of the process described above or only of the certain phases within it with purpose to identify objectives for a new simulation or to achieve the objectives which have not been achieved. The advantages of using SD and ABMS models Advantages of using the SD models are reflected by the fact that this methodology allows the modeling of the enterprise behavior at top-level decision-makers through the capture of the dynamic system at the macro level. SD provides an overview of the system and use causal diagrams to represent the model created. Another advantage is that the simulation SD can also be used as a model for decision support, because it allows parameters and performances model adjustment through a real time analysis. The causal structure of the model allows decision-makers to quickly identify key variables and decision-making options that can be used to control the behavior of the system. ABMS represents the only simulation methodology capable for rational behavior shaping at the locally level based on bottom-up approach in the context of systems with local entities distributed. ABMS is also a flexible, efficient and adaptable methodology which by using artificial intelligence allows bottom-top capturing of the dynamics generated by the interactions of autonomous agents and the way in how they respond to the environment. Due to the characteristics they possess (autonomy, self-orientation, precise direction toward a certain goal, flexibility) agents can be modeled with a high fidelity and without rigid assumptions as in mathematical modeling which allows a realistic representation of the actors responsible for decision-making processes (Sawyer 2003). ABSM can be used in different contexts and allows modeling of systems composed in both homogeneous and heterogeneous entities. Classes of agents with different characteristics can interact within the same model because they are designed to self-evolving and to selfadjusting structure during the model execution. Agents are used for exploring space problems by developing a exploration structure on the basis of schemas incorporating intelligent heuristics algorithms search (Scheritz and Größler 2003). This capability allows simulations creation without a prior knowledge of the system macro-structure. Agents also integrate the heuristics and genetic operators used for uncertain environments exploring in order to new objectives discovering. Based on micro to macro level approach, the methodology enables both hypothesis and different scenarios testing and different rules and parameters establishing (Guyot and Honiden, 2006).

72

www.hrmars.com

International Journal of Academic Research in Economics and Management Sciences May 2013, Vol. 2, No. 3 ISSN: 2226-3624

The limitations of using SD and ABMS models The main limitation of SD methodology is reflected by the fact that it uses only aggregated terms in structure and behavior modeling – all people and resources are treated as a single homogenous resource that vary continuously. This issue can be approached as an advantage for the model computerization, but also as a disadvantage in applying the model to simulate heterogeneous systems. Because the terms aggregation design facilitate implementation of systems with a high level structure, it is recommended that the SD to be used for modeling dependencies and dynamics of enterprise macro-level. A major disadvantage of ABMS is represented by the high difficulty to verify and to validate the behavior models simulated and the relationship between the agents micro level schema and macro system resulted behavior. According to Rahmandad and Sterman (2004) the difficulty of the model testing and evaluation is generated by the use of a high number of agents which requires increasing the number of parameters used. Another limitation of ABMS model consist on the fact that its methodology cannot be applied to those organizational problems with routine and non-deterministic character (like processes) which do not exhibit a top-down behavior and to organizational issues such as computer systems, processes, strategic planning and resource allocation. In this case processes modeling and simulation can be done better using DS methodology. Because ABMS is based on the complex schemes that include memory and artificial intelligence, these models can become very complicated and computer resources intensive consumers compared to other simulation techniques and models. Conclusions Based on the previous discussed issues, we consider that although each methodology captures a certain dimension of system behavior, none is able to capture all aspects that characterize the enterprise complex behavior. In this context, the selection of a suitable methodology can be achieved depending on the nature of the problem to be simulated and considering the advantages and the limitations of specific methodologies, and also the factors such as analytical ability, difficulty of implementation and ease of communication to stakeholders and to the model end-users. ABMS methodology is recommended for modeling systems with the micro to macro level behaviors or to those which require the model adaptation during simulation. For many macro level systems, the selection methodology is still unclear, although the most viable approach is considered to be intelligent ABMS and SD equations based.

73

www.hrmars.com

International Journal of Academic Research in Economics and Management Sciences May 2013, Vol. 2, No. 3 ISSN: 2226-3624

Although there are no specific tests for determining the optimal methodology for simulation of a particular issue or its dynamics, this article can provide a guide for choosing the appropriate simulation approach for obtaining a correct analytical result. Due to differences in the simulation methodologies and to limits in individual uses, recent researches in this field aim to study the opportunity of combining them and to develop hybrid simulation models. Acknowledgments This work was co-financed from the European Social Fund through Sectorial Operational Programme Human Resources Development 2007-2013; project number POSDRU/107/1.5/S/77213 „Ph.D. for a career in interdisciplinary economic research at the European standards”. References 1. Akkermans, H. (2001). Emergent supply networks: System dynamics simulation of adaptive supply agents. In Proceedings of the 34th Hawaii International conference on system science, January 3–6, Maui, USA. 2. Anderson, P., A. Meyer, et al. (1999). "Introduction to the Special Issue: Applications of Complexity Theory to Organizational Science." Organization Science 10 (3): 233-236. 3. Angerhofer, B.J., Angelides, M.C., (2006). A model and a performance measurement system for collaborative supply chains. Journal Decision Support Systems, 42(1), 283301 4. Axelrod, R., & Tesfatsion, L., (2006). A Guide for Newcomers to Agent-Based Modeling in the Social Sciences, Staff General Research Papers 12515, Iowa State University, Department of Economics 5. Burton, R. M. ,(2003). Computational laboratories for organization science: Questions, validity and docking. Computational and Mathematical Organization Theory, 9, 91–108. 6. Carley, K. M. (2002). Computational organizational science and organizational engineering. Simulation Modelling Practice and Theory, 10, 253-269. 7. Cohen, K. J., & Cyert, R. M. ,(1965). Simulation of organizational behavior. In J. G. March (Ed.), Handbook of organizations: 305–334. Chicago: Rand McNally. 8. Ford, D.N., Sterman, J.D., (1998), Dynamic modeling of product development processes. System Dynamics Review, 14(1), 31-68 9. Forrester, J. W. (1958). Industrial dynamics – a major breakthrough for decision makers. Harvard Business Review, 36(4), 37–66. 10. Fowler, A. (2003). "Systems Modelling, simulation, and the dynamics of strategy." Journal of Business Research ,56, 135-144. 11. Garro, A., Russo. W., (2010). EasyABMS: A domain-expert oriented methodology for agent-based modeling and simulation. Simulation Modelling Practice and Theory, 18 ,1453–1467 74

www.hrmars.com

International Journal of Academic Research in Economics and Management Sciences May 2013, Vol. 2, No. 3 ISSN: 2226-3624

12. Guyot, P. and S. Honiden (2006). "Agent-Based Participatory Simulations: Merging MultiAgent Systems and Role Playing Games." Journal of Artificial Societies and Social Simulation 9(4): 8. 13. Hilletofth, P., Aslam, T., & Hilmola, O.P. (2010). Multi-agent based supply chain management: Case study of requisites. International Journal of Networking and Virtual Organisations, 7(2/3), 184–206. 14. James M Lyneis, J.M., (2000). System dynamics for market forecasting and structural analysis. System Dynamics Review, 16(1), 3-25 15. Kambiz E. Maani, K.E., Maharaj, V., (2004) Links between systems thinking and complex decision making.System Dynamics Review, 20(1), 21-48 16. Keskin, B.B., Melouk, S.H., Meyer I.L., (2010). A simulation-optimization approach for integrated sourcing and inventory decisions. Computers & Operations Research, 37, 1648–1661 17. Lättilä,L.,Hilletofth, P., Lin, B.(2010). Hybrid simulation models – When, Why, How? Expert Systems with Applications, 37, 7969–7975 18. Macy, M. W., & Willer, R. , (2002). From factors to actors: Computational sociology and agent-based modeling. Annual Review of Sociology, 28, 143–166. 19. Miragliotta,G.,Brun,A.,Soydan,I.A.,(2009). Coordinating multi-business sales through management simulators. Int. J. Production , 121, 533–549 20. Nilsson & Darley (2006). On complex adaptive systems and agent-based modelling for improving decision-making in manufacturing and logistics settings – Experiences from a packaging company. International Journal of Operations and Production Management, 26(11), 1351–1373. 21. Rahmandad, H., & Sterman, J. (2004). Heterogeneity and network structure in the dynamics of contagion: Comparing agent-based and differential equation models. In Proceedings of the 22nd international conference of the system dynamics society, July 25–29, Oxford, England. 22. Repenning, N. P. 2002. A simulation-based approach to understanding the dynamics of innovation implementation. Organization Science, 13, 107–127. 23. Sawyer, R. K. (2003). "Artificial Societies: Muliagent Systems and the Micro-Macro Link in Sociological Theory." Sociological Methods & Research 31(3): 325-363. 24. Schieritz, N., & Grössler, A. (2003). Emergent structures in supply chains – A study integrating agent-based and system dynamics modeling. In Proceedings of the 36th Hawaii International conference on system sciences, January 6–9, Waikoloa, USA. 25. Scholl, H. J. (2001). Agent-based and system dynamics modeling: A call for cross study and joint research. In Proceedings of the 34th Hawaii international conference on system sciences, January 3–6, Maui, USA. 26. Senge, P.M., (1990). The Fifth Discipline: The Art and Practice of the Learning Organisation. Random House, London. 27. Sterman, J. D. (2000). Business dynamics: Systems thinking and modeling for a complex world. United States: McGraw-Hill. 28. Swinerd, C., McNaught, K.R., (2012), Design classes for hybrid simulations involving agent-based and system dynamics models. Simulation Modelling Practice and Theory, 25, 118–133 75

www.hrmars.com