Simulations are imitations of some “real-world” phenomena, especially the states of affairs of (real) natural or social systems or the processes of the systems (processes are defined as temporal sequences of system states). Simulations are used in numerous domains: physics, chemistry, biology, economy, social sciences, and computer and engineering sciences. The general purpose of simulations is to attain insight into the functioning of these systems and to predict their future states.
The main meaning of simulation in the social sciences, communication, cognitive science, and artificial intelligence is computer simulation (CS). Simulations in this sense should always be defined with reference to a “real-world” phenomenon and to an abstract (theoretical) model. If a researcher is interested in a phenomenon (the target) he or she often tries to create an abstract model that is less complex than the target itself. In the social sciences the target usually is a dynamic entity. It changes over time and reacts to its environment. Therefore the model must capture the dynamics of the target. The model can be specified, for example, via a logical calculus or a mathematical equation connecting relevant variables. The development of the model over time can then be explored by the inspection of the development of its relevant variables. In the case of complex models, especially with nonlinear relations (e.g., nonlinear ordinary differential equations) and/or many interacting subsystems this can only be done with the aid of CS.
Simulation now means running the specified model forward through time and inspecting future states of the relevant variables. CS can be helpful if analytical methods are also available. Visualizing simulation results on a computer screen often increases the understanding of a system’s dynamics compared to complicated formulae and their results written down on paper. To summarize: scientific simulations are experiments with an abstract model to gain a better understanding about the dynamic system selectively represented in this model. Famous early examples of simulations are the world models of the Club of Rome based on the method of system dynamics developed by Forrester (1971).
Simulations as scientific tools serve various functions (Hartmann 1996). As a technique they allow the investigation of the dynamics of systems. But simulations can also serve an important function as a heuristic tool in the process of the formulation of hypotheses and new theories. The analysis of huge sets of runs with systematically or randomly varied parameters may yield unexpected regularities that would not have been inferred from the underlying abstract model. Furthermore, simulations can substitute and extend experiments. If the performance of an experiment is impossible because the real system is withdrawn from human access (e.g., in astronomy), or because an experiment would cause unreasonable modifications of the system or is morally not permissible (e.g., children’s long-term viewing of aggressive contents), the only possibility of testing abstract models remains in simulations. Simulations are extensions of experiments, because not only can the effect of a specific experimental design be measured but the entire universe of possible condition variations can be inspected. Therefore simulations can be used as generators of ideas for condition variations in laboratory experiments.
One special form is simulations with multi-agent systems (MAS), which are used to model the dynamic behavior of populations (Wooldridge 2002). An MAS is a system running on a computer and composed of several agents and their environment. The definition of agents is still controversial, but in MAS an agent mostly represents a life-like artificial unit. Agent-based simulations are suited to models and explain the emergence of higher-level phenomena caused by the interaction of single agents. The interaction of many agents on a microscopic level will sometimes result in stable systems on a macroscopic level (e.g., the emergence of communication systems) that show novel and sometimes unexpected properties. Compared to traditional simulation techniques, MAS have the advantage that global behaviors as well as each agent’s individual behavior, including its causes and consequences, can be observed. Examples of social phenomena simulated with MAS are the emergence of social intelligence, the emergence of intergroup cooperation, the spreading of gossip, and the evolution of communication, to mention just a few.
One special case of simulations is interactive or human-in-the-loop simulations (HILS). These differ from other simulations in that human operators or users are included in the simulation process. Examples of HILS are medical simulators, flight simulators, and driving simulators. In most cases the goal of HILS is not simply the prediction of future states of dynamic systems. HILS are often used as experimental settings in human– computer interaction research to measure user behavior patterns. The results of the simulations can be used to improve the usability of interfaces (of cars or airplanes). Another function of HILS is the training of personnel. If it is too dangerous or too expensive to permit trainees the use of real equipment, learning in a safe virtual environment is a preferable alternative. To test the fitness of pilots, flight simulators can be used to simulate extremely hazardous scenarios.
Almost all HILS are realized not only as training devices but also as video and computer games. City simulators are tools for urban planners to gain a better understanding of the evolution of a city, but they are also the game engines for computer games. Simulations form a category of their own in the domain of video games. This category can be further subdivided into sensorimotor simulations (e.g., flight and racing simulations) and simulations that can be used by the player to organize a dynamic system like a family, an organization, an economic system, or the evolution of a species. Massively multi-player online role-playing games (MMORPGs) are very complex simulations of alternative virtual worlds that have no counterparts in the domain of scientific simulations. Virtual environments like Second Life are also simulations with comparable complexity, but they are not games because they do not have a pre-defined goal that the user is expected to reach.
References:
- Forrester, J. W. (1971). World dynamics. Cambridge, MA: Wright-Allen.
- Hartmann, S. (1996). The world as a process: Simulations in the natural and social sciences. In R. Hegselmann, U. Mueller, & K. G. Troitzsch (eds.), Simulation and modelling in the social sciences from the philosophy of science point of view. Dordrecht: Kluwer, pp. 77–100.
- Wooldridge, M. (2002). An introduction to multiagent systems. Chichester: John Wiley.