The term model is used in multiple ways in science and there are several different kinds of models. The most basic scientific models are material and conceptual analogues. They are copies that stand in for more opaque systems. Cloud chambers and cell cultures are examples of material models, whereas conceptual models are more abstract analogies that seek to render theories more comprehensible. Mathematical models are typically applications, approximations, or specifications of theories and principles that cannot be applied in their original form.
Computer simulations tend to be more obvious analogues than models, aptly characterized as virtual copies of systems. Simulations employ a generative mechanism to imitate the dynamic behavior of the underlying process that the simulations aim to represent. Simulation is an ambiguous term, but in all cases of scientific simulations they are based on some form of model. However, simulation models can be divided into two overall types, both of which tend to be used for representations of complex dynamics. The first type is based on mathematical models, which aim to represent established theoretical statements or physical laws. Simulations in physically based sciences usually exemplify this type of simulation. The second type of simulation model is based on simpler models, which consist of a few assumptions about leading mechanisms. This is generally the case in simulations of social phenomena using so called agent based models.
Although scientific models and simulations are increasingly used in science, studies of them remain rare compared to the extensive studies of experiments that have taken place in the sociology of science since the 1970s. Models and simulations that represent the application of theoretical knowledge have also received less attention in the philosophy of science, where the traditional line of inquiry has mostly been in the theoretical domains.
From the perspective of the philosophy of science, models and simulations tend to be discussed in relation to theories. In the classic contribution Models and Analogies in Science (1966), Mary Hesse sees models as heuristically essential to the development and extension of theories, and also essential to the explanatory power of theories. However, the epistemological characteristics of models and simulations are less clear than those of theories. Models do not have the same epistemic tradition as theorizing and do not have transparent object domains. This is perhaps most significant in the case of simulations and it has thus been argued that simulation has its own epistemology (Winsberg 1999). In constructing simulation models, theoretical structures are transformed into specific knowledge of systems and further into computational models that are implemented in a computer in the form of an algorithm. This is what makes simulations produce data sets. Because simulation modeling produces these types of results, a standard for deciding whether the results are reliable is required. For this reason, results of simulations are often compared to observations. In this respect modeling and simulations share similar relations to experimentation as does theory. However, models and simulations are often seen in more pragmatic terms and therefore evaluated in relation to the purpose of modeling. This approach differs clearly from the epistemological view on theories, where theories are judged according to their being true or false.
A different way to approach modeling, not limited to epistemic questions such as the truth of models and how they should be verified, is to consider the instrumentality and autonomy of models. As partly dependent on theories and experiment, and partly independent, models can serve as bridges between theory and the world. This is suggested by Mary Morgan and Margaret Morrison in their influential book Models as Mediators (1999), where models are conceptualized as autonomous mediators. However, to only see models and simulations in the space between theories and the world assumes that theory and the world are stable entities, at the same time as it directs the attention toward a philosophical focus on what a model essentially is. From a sociological perspective, it is more appropriate to address the character of models in relation to the role that models and simulations play in scientific practice.
The characterization of the role of models and simulation models has often been based on the idea of models and simulation models as tools or objects for knowledge. The use of models and simulations as tools is most evident in applied science, where models are used to predict the development of various dynamic processes. Thus, in those cases where empirical data exist, the correspondence of outcomes with data becomes an important indicator of the performance of the model or simulation.
Because of their character as analogues, models and simulations can also themselves be studied the same way as natural systems are studied in empirical research. By acting as objects of knowledge in their own right, simulation models are explored to answer questions about how and why certain processes develop. In this situation, both the inner theoretical structure of the model and how well its results correspond to data are important to the evaluation of the performance of the simulation model itself.
Some models and simulation models are constructed and used for only one of the above mentioned purposes. However, in principle, a particular simulation model can serve multiple purposes, depending on its role in practice and which questions it is being asked to address. Consequently, some models and simulations may serve as tools like technical artifacts or objects of knowledge depending on the setting where they are used.
Another line of research focuses on what people who model and simulate do, and how they work. While simulations can be described as both experimenting and theorizing depending on what aspects researchers talk about, the use of simulations as ”virtual laboratories” in fact makes working with models very similar to experimenting (Dowling 1999). Models tend to integrate a broad range of ingredients such as, for example, metaphors, theoretical notions, and mathematical concepts and, not least from this point of view, the construction of models requires much experience and hard work (cf. Boumans 1999). In simulation modeling, questions related to the role of researchers are fundamental because the construction of simulation models and the interpretation of simulation results primarily depend on the researchers, their research areas, and their experience (Becker et al. 2005). In short, the role of human agency needs to be taken into account in developing the sociological understanding of how models and simulation models are constructed and used.
What appears as a particularly useful and important way to approach the practice of modeling and simulation is to acknowledge the materiality of modeling and simulations. Models are objects that have their own construction and ways of functioning that constrain interpretation and use (Knuuttila & Voutilainen 2003). What makes the materiality of simulation modeling even more evident is the transformation of a theoretical model into a computer program, and this intertwining is indeed a fundamental aspect to attend to in understanding simulation modeling practice (cf. Sundberg 2005).
However, further exploration is needed in terms of how the practices of modeling and simulating can be conceived of in different ways rather than only in relation to theorizing and/or experimenting. In addition, an important question for future research concerns whether the concepts, metaphors, and methodologies developed on the basis of studies of experiments and experimental work can be successfully applied to and used in studies of modeling and simulations. For example, the participant observation approach, which has been the basis of many so called laboratory studies, is more difficult when studying modelers. Compared to the traditional work in a ”wet laboratory,” it is more difficult for an observer to follow activities like writing equations or programming computers. To conclude, there is a growing but limited interest in, and knowledge of, the practices involved in modeling and simulations, but a more rigorous sociological approach remains to be developed.
References:
- Becker, J., Niehaves, B., & Klose, K. (2005) A Framework for Epistemological Perspectives on Simulation. Journal of Artificial Societies and Social Simulation 8(4).
- Boumans, M. (1999) Built-in-Justification. In: Morgan, M. S. & Morrison, M. (Eds.), Models as Mediators: Perspectives on Natural and Social Science. Cambridge University Press, Cambridge.
- Dowling, D. (1999) Experimenting on Theories. Science in Context 12(2): 261-73.
- Knuuttila, T. & Voutilainen, A. (2003) A Parser as an Epistemic Artifact: A Material View on Models. Philosophy of Science 70: 1484-95.
- Lenhard, J., Kuppers, G., & Shinn, T. (Eds.) (2006) Simulation: Pragmatic Constructions of Reality. Sociology of Sciences Yearbook. Springer, New York.
- Sundberg, M. (2005) Making Meteorology: Social Relations and Scientific Practice. Acta Universitatis Stockholmiensis. Stockholm Studies in Sociology, N.S. 25. Almqvist & Wiksell, Stockholm.
- Winsberg, E. (1999) Sanctioning Models: The Epistemology of Simulations. Science in Context 12(2): 275-92.
- Zeigler, B. (1976) Theory of Modeling and Simulation. Krieger, Malabar.
Back to Sociology of Science
