Tools or Toys?
|
| Table of Contents |
As philosophers that take the “simulations are experiments”-side I would just like to quote Margarete Morrison and Uskali Mäkki. Morrsion writes in an article titled “Models, measurement and computer simulation: the changing face of experimentation”:
hence we have no justifiable reason to assume that, in these types of cases, experiment and simulation are methodologically or epistemically different. As we shall see, the causal connections between measurement and the object/property being measured that are typically invoked to validate experimental knowledge are also present in simulations. Consequently the ability to detect and rule our error is comparable in each case. (Morrison 2009, p.\ 43/44) [The conclusion, that simulation can attain an epistemic status comparable to laboratory experimentation, involved showing its connections with particular types of modelling strategies and highlighting the ways in which those strategies are also an essential feature of experimentation. (Morrison 2009, p.\ 55/56)
Interestingly, Morrison concedes in a footnote that the account of computer simulations she has argued for may be more appropriate for the natural than the social sciences (Morrison 2009, p.\ 56, fn 33).
In a similar vein, Uskali Mäkki compares experiments with models and reaches the conclusion that “Models are Experiments, Experiments are Models” (Maeki 2005). And the same holds - as we may add without distorting his idea - a fortiori for computer simulations:
Consider material experimentation as based on causally isolating fragments of the world from the rest of it so as to examine the properties of those fragments free from complications arising from the involvement of the rest of the world. The analogy with theoretical modelling is obvious: while material experimentation employs causally effected controls, theoretical modelling uses assumptions to effect the required controls. (Maeki 2005, p.\ 308)
Maeki, however, does not claim that models and experiments can always be equated in this way. And he sees a difference between models and experiments in the fact that in the case of experiments the isolation of causal factors requires material manipulation. A difference that Morrison, in contrast, considers to be rather inessential (Morrison 2009, p.\ 54).
That the view that simulations are experiments is not merely an eccentric philosophical point of view is further illustrated by the fact that practitioners often use the term “simulation experiments” as a facon de parler for referring to computer simulations. For example, Hegselmann and Flache use the term “experiment” when referring to results obtained with cellular automata that run on the computer (Hegselmann/Flache 1998, 3.11). And in a recent simulation of customer experience in retail stores, Siebers, Aickelin, Celia and Clegg use the terminology of hypothesis and experiments when referring to pure simulation studies (Siebers et al. 2010, p.\ 16ff.). They are aware, however, that these kinds of “experiments” are not empirical - in contrast to what they call the “validation experiments” of their simulation.
The opposite position with regards to the simulation-experiments question is, among many others, taken by Kleindorfer and Ganeshan, who place simulations firmly on the theoretical side of science by declaring them with reference to Naylor and Finger (Naylor/Finger 1967) as “miniature scientific theories”:
To simulate means to build a likeness and the question as to the accuracy of the likeness, one version of the validation problem (some might argue the only version), is never far behind. The validation problem is an explicit recognition that simulation models are like miniature scientific theories. (Kleindorfer/Ganeshan 1993, p.\ 50)
If simulations are understood in analogy to theories rather than experiments it appears only natural to also consider the validation requirements of simulations in analogy to that of theories. This is what Troitzsch does when he maintains that “Validation of simulation models is thus the same (or at least analogous) to validation of theories.” (Troitzsch 2004)[3] And Naomi Oreskes, Kristin Shrader-Frechette and Kenneth Belitz even warn that “Any scientist who is asked to use a model to verify or validate a predetermined result should be suspicious.” (Oreskes et al. 1994, p.\ 644) Given that one denies that simulations are experiments this warning is important, because then, rather than being able to “verify or validate” given assumptions, models and simulations are in the need of verification and validation themselves.
|
[3] See also K_uuml_ppers (2005), where I took these references from, in this context.
|