Abstract
Computer simulations have conventionally been understood to be either extensions of formal methods such as mathematical models or as special cases of empirical practices such as experiments. Here, I argue that computer simulations are best understood as instruments. Understanding them as such can better elucidate their actual role as well as their potential epistemic standing in relation to science and other scientific methods, practices and devices.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Notes
Here I treat the definition of an instrument as being maximally tolerant and as admitting of many functional technical artifacts. For the purposes of this paper, and following Davis Baird (2004), what is required of this view so far is to say that whatever instruments are, they are simply not propositions, nor are they necessarily constituted by propositional knowledge. Instruments are something else than propositions alone. And while they must be cases of instantiated design, they need not exercise causal interventions on the subject of inquiry., i.e. they can be measuring devices and detecting devices without necessarily coming into contact with the object of inquiry or intervening in a conventionally understood empirical manner. This, however, is different from a scientific instrument which on top of the minimum requirements above must also undergo many rigorous tests and the scrutiny of superior epistemic requirements. Whether or not computer simulations qualify as the latter is precisely the question and conversation whose necessary groundwork this paper seeks to establish for future research. Importantly, as we will see in Sect. 3 of this paper, instruments can be comprised of other instruments. I thank an anonymous reviewer for emphasizing the need for a definition of the term ‘instrument’.
As we will see below, of particular interest to this investigation are the technical artifacts whose functionality is that of enhancing our epistemic abilities: epistemic enhancers (Humphreys, 2004)—objects deployed in the aid of inquiry with the aim of enhancing knowledge or our capacities to acquire it. This is important because we must acknowledge that the general definition of instruments provided above—which so far makes no mention of scientific instruments—allows for the existence of technical artifacts that can diminish capacities rather than enhance them (i.e. instruments of torture, resistance training instruments, etc.). Few if any of these play a significant role in scientific inquiry. So, the category of interest here is significantly smaller than the category of general objects that can be identified as instruments in the general sense.
While a description of an instrument may meet this criteria notice that the description qua description is a different kind of instrument than the instrument it describes. A blueprint of a technical design may enhance our understanding of the thing it represents, but it is not the things it represents nor does it do the same things as the instantiation of the thing it depicts.
Even without the complexities introduced by computational methods, the history of the notion of instrument in scientific inquiry has its ambiguities. The term, for example, was not used before the 1800’s (Warner, 1990). Before then, the devices surrounding what we now understand as practical methods of hypothesis demonstration were called philosophical instruments, whereas those associated with measuring were called mathematical instruments. A discussion of the role of devices in serious inquiry can be found discussed as early as 1649 in correspondence between Boyle and Bacon (Warner, 1990 p. 83). The key distinction here is between the kinds of artifacts deemed to furnish information about the world and those that were deemed to be merely a product of an arbitrary formal construct. The distinction between philosophical instruments and ‘mere’ mathematical instruments is similar to the distinction between those instruments that have direct, what we now understand as empirical, access to an aspect of the world and those that are constituted of and solely manipulate mere formal conventions.
While there are many issues regarding this intuition for the purposes of the present discussion, this is a good-enough way to tell objects apart.
Kroes calls the former pseudo-artifacts in virtue of the fact that they are only circumstantially artifactual and not intentionally so.
As the discussion in Sect. 3.1 will show, according to Baird, this epistemic independence of instruments derives from the fact that some uses of instruments can both precede theoretical input as well as fail to partake in conventional empirical interventions. He offers the physical construction of the double helix model of DNA by Watson and Crick. Watson and Crick, according to Baird, constructed and used this model without having much in the form of theoretical foundations to do so. Furthermore, the model also seemed to provide insight into the nature of DNA without having had any direct causal manipulation of the target phenomenon.
Of course, particularly in science, this independence can be seen as purely conceptual. The role instruments play in scientific inquiry is strongly informed by theoretical principles as well as by experimental aims. But this does not take away from the fact that they are unequivocally not identical to one or the other.
It is worth stating at this point that important philosophical contributions have been made in the last two decades regarding the non-trivial ways in which computer simulations are distinct from purely formal methods as well as from conventional empirical practices such as laboratory experiments. I want to thank the anonymous reviewer who rightly emphasized this point. However, often, these distinctive features tend to be used for a different inference than the one in this paper. They are used to say that computer simulations represent a special case within the same category they are being distinguished from. For example, if the contrast is between the distinctive character of computer simulations from that of models, this distinctiveness is then used to argue that computer simulations are therefor a special case of models (see Weisberg, 2012 for why this is the case). The same happens when computer simulations are being contrasted to experiments (see Barberousse et al., 2009; Morrison, 2015). Their distinctiveness is used to infer that they must be a special kind of experiment. In this paper, no such inference takes place. Rather, in this section the main argumentative point is to show that whatever else they may be, computer simulations are no the things the conventional debate deems them to be. Furthermore, in following sections, this point is used in an argument to the best explanation claiming that computer simulations are best understood as instruments. This is because their in-betweenness is best captured by this framework, but also because of the particular unificatory virtues of this view.
It is important to note that digital implementations also fit this description.
A simulation pipeline is comprised of several stages. Winsberg’s (2010) description of the pipeline includes starting out with a theory of an observed phenomenon, followed by a model, then a treatment of the model, and then a solver that yields results. Resch’s (2013; 2017) pipeline description breaks down the process further the following way: (1) observation of reality, (2) abstraction of physical model, (3) conceptualization of mathematical model, (4) translation of 3 into a numerical scheme, (5) translating 4 into a program structure, (6) designing a programming model for 5, and (7) engineering a hardware architecture.
The reader need not at this point be convinced of what computer simulations are to accept what they are not. Hence, even if the reader remains unconvinced that any of the discussion above points towards the fact that computer simulations are instruments, the observation that they are something distinct from any one of the stages of a simulation pipeline or distinct from the results arrived at by the processes of such pipeline still holds.
Often, even the necessary idealized or otherwise false abstractions in scientific mathematics are theoretically justified in relation to the phenomena they seek to capture (for more on the role of necessary false abstractions in scientific mathematics see Pincock, 2015).
While there is a substantial amount of work in the philosophy of science concerning the nature of experimental practices, to engage with these debates is beyond the scope of this paper. What is meant here is simply that a simulation of an experiment is not identical to the experiment itself or to the description of the experiment. It is not identical because the computer simulation’s function is to simulate the otherwise existing experiment. Hence, it is doing something else, namely simulating that experiment.
While some elements of Kroes’ nomenclature involve sociological considerations such that a term like technical artifact may actually be meant to be understood as a sociotechnical artifact, Symons’ idea that intended functions are fundamental to the identification of artifacts already implies the possibility of this identification hinging on sociological factors. For the purposes of this paper, the metaphysical claim observing that there is simply a distinction between objects found in nature and intentionally synthesized resources may suffice.
While some may point out that a computer simulation can be the experiment itself, I take this to mean that the computer simulation can become the subject of inquiry of an experimental set up. This, again, is different from saying that the computer simulation is the experiment. More precisely, it is not the case that there is an identity relation between what they are calling the experiment and the computer simulation: most of the time in these settings, the computer simulation is either the subject of inquiry of the experiment or the thing that carries out an experiment and therefore not identical.
I am grateful for the anonymous reviewer that encouraged me to emphasize this point. While it is true that important and extensive work has been done highlighting the distinctive features that separate computer simulations from both abstract models and empirical experimentations, the nature of the arguments, the reasons and the conclusion of such accounts is vastly different from the perspective in this paper. In particular, the conceptual and functional distinctions of computer simulations that draw from the metaphysics of artifacts (Symons, 2010) are not at all in the literature. Most arguments for the distinctiveness of computer simulations conclude that they are special cases within a set dichotomy between theoretical abstractions and empirical practices rather than ontological and hence epistemically distinct objects. The main point of this paper is the latter.
A very significant element of the practice of modeling which is often conceived as the task of creating a computer simulation in practitioner’s circles consists of testing the adequacy, efficiency and overall viability of many mathematical models of a target phenomenon.
This is what Lenhard (2007) referred to as the ‘quasi-empirical’ aspect of simulation modeling and what Morrison (2015) meant when she wrote that computer simulations allowed for ‘testing’ of both theoretical and experimental stipulation. Herbert Simon (1996) too, identified the potential for computer simulations to not only provide means to solve an equation but also provide the means to explore the numerous ways in which an equation could be solved.
There may also be a difference between those artifacts that enhance the means by which we gain understanding and those that directly enhance our understanding, though that is a topic far too removed from the discussion in this work.
There is an important philosophical question concerning what exactly a telescope allows us to do. That is, what it is that telescopes allow us to enhance. For a thorough discussion of whether and how we can actually ‘look closer’ with a telescope see Hacking and Hacking (1983).
Humphreys does not believe that computer simulations have the capacity to enhance our capabilities in mathematics: as explained in a footnote earlier, according to Humphreys, computer simulations solve intractable problems in mathematics. They do this by being faster than us but not by providing us with a novel way of doing mathematics (representational opacity notwithstanding). Therefore, if they are epistemic enhancers, they are so only in the sense that they can calculate faster than humans. I find this claim to be somewhat misguided. Novel representational devices have in the past enhanced our access to areas of knowledge not previously available to us. Consider the notation of calculus, or the advent of statistical concepts such as the average which allowed us to gain previously inaccessible phenomenal insights thanks in particular to a novel manner of symbolic manipulation and aggregation. Furthermore, statistical reasoning expanded the kinds of things we could know about a phenomena, such as the evolution of galaxies, that we are not naturally equipped to access. In this sense, the representational capacities of computer simulations may very well be positioned to provide us access to aspects of the world previously unavailable to us, and these aspects may prove to be mathematical at their foundation even though not immediately apparent to us as such. Here I am thinking of neural networks and other similar computational methods in machine learning and statistical analysis, which may indeed reveal to us relationships and patterns that would otherwise be unavailable or unimaginable.
Symons and Boschetti focus is on the predictive function of computational models and hence their epistemic import: what do they do and what does this contribute to a knowledge gathering endeavor such as scientific inquiry? It is unclear, however, whether Symons and Boschetti refer to computational models proper or computer simulations more generally. While they refer throughout their paper to computational models, their references to Parker (2009), Frigg and Reiss (2009), and Winsberg (2010) (pp. 810) point to the fact that what they have in mind are computer simulations.
As explained in the introduction, the argument here for the epistemic independence of instruments is fairly straightforward: according to Baird, the knowledge gained by Watson and Crick from the model was knowledge that was not acquired via nor implied by either the use of theoretical principles (there weren’t any to draw from) or from conventional experimental procedures i.e. interventions with the actual target of inquiry. Hence, the knowledge was independent from both.
Due to limited space, I have omitted a discussion on the important differences between my view of instruments and technical artifacts and by extension my view of computer simulations and that of Harré’s (2003), particularly as it related to his taxonomy of instrumentation. I want to thank an anonymous reviewer for suggesting a closer look of Harré’s position. In Harré’s view, things in the laboratory can be an apparatus, “an arrangement of material stuff integrated into the material world in a number of different ways”—or an instrument “which registers an effect of some state of the material environment.” (2003 p. 19) If seen as conditions, i.e. requirements that must be satisfied to belong to one category or another, there is a sense in which computer simulations as described above could fail to meet them. Of course, to deny that they meet the overly broad first condition—that they must be arrangements of material stuff integrated into the material world—would require a discussion that is beyond the scope of this paper. What can be said for now, is that computer simulations are carried out by arrangements of material processes, they are instantiated in the world, they have functions that they carry out and they require the instantiation of specific procedures. As for meeting the second condition—that it register an effect of some state of the material environment, things will depend on what is meant by ‘register’, ‘effect’ and ‘some state of the material environment’. On a narrow interpretation, computer simulations do not register effects directly from the material environment. They are not detecting mechanisms such as sensors. They are different from an analog thermometer. Particularly due to their computational mediation. However, in a broader interpretation, computer simulations may not be as distinct from a digital thermometer which is a software-intensive (Symons & Horner, 2014) instrument.
References
Alvarado, R. (2020). Computer simulations as scientific instruments (Doctoral dissertation, University of Kansas).
Baird, D. (2004). Thing knowledge: A philosophy of scientific instruments. University of California Press.
Barberousse, A., & Jebeile, J. (2019). How do the validations of simulations and experiments compare? In C. Beisbart & N. Saam (Eds.), Computer Simulation Validation. New York: Springer.
Barberousse, A., & Vorms, M. (2014). About the warrants of computer-based empirical knowledge. Synthese, 191(15), 3595–3620.
Dalmedico, A. D. (2001). History and epistemology of models: Meteorology (1946–1963) as a case study. Archive for History of Exact Sciences, 55(5), 395–422.
Durán, J. M. (2018). Computer Simulations in Science and Engineering. Springer.
Durán, J. M. (2019). A formal framework for computer simulations: Surveying the historical record and finding their philosophical roots. Philosophy & Technology, 34, 105–127.
Durán, J. M. (2020). What is a simulation model? Minds and Machines, 30(3), 301–323.
Frigg, R., & Reiss, J. (2009). The philosophy of simulation: Hot new issues or same old stew? Synthese, 169(3), 593–613.
Gransche, B. (2017). The art of staging simulations: Mise-en-scène, social impact, and simulation literacy. In The Science and Art of Simulation I (pp. 33–50). Cham: Springer.
Hacking, I., & Hacking, J. (1983). Representing and intervening: Introductory topics in the philosophy of natural science. Cambridge university press.
Harré, R. (2003). The materiality of instruments in a metaphysics for experiments. In H. Radder (Ed.), The Philosophy of Scientific Experimentation. Pittsburgh: University of Pittsburgh Press.
Heidelberger, M. (2003). Theory-ladenness and scientific instruments in experimentation. In H. Radder (Ed.), The Philosophy of Scientific Experimentation. Pittsburgh: University of Pittsburgh Press.
Heilbron, J. (1993). Some uses for catalogues of old scientific instruments. In R. G. W. Anderson, J. A. Bennett & W. F. Ryan (Eds.), Making instruments count: Essays on historical scientific instruments presented to Gerard L’Estrange Turner. Brookfield: Variorum.
Helden, V., & Albert, and Hankins, Thomas L. (1994). Introduction: Instruments in the history of science. Osiris, 9, 1–6.
Humphreys, P. (2004). Extending ourselves: Computational science, empiricism, and scientific method. Oxford University Press.
Humphreys, P. (2009). The philosophical novelty of computer simulation methods. Synthese, 169(3), 615–626.
Imbert, C. (2017). Computer simulations and computational models in science. In Springer handbook of model-based science (pp. 735–781). Cham: Springer.
Keller, E. F. (2003). Models, simulation, and “computer experiments”.
Kroes, P. (2002). Design methodology and the nature of technical artefacts. Design Studies., 23(3), 287–302.
Kroes, P. (2003). Physics, experiments, and the concept of nature. In H. Radder (Ed.), The philosophy of scientific experimentation. University of Pittsburgh Pre.
Kroes, P. (2006). Coherence of structural and functional descriptions of technical artefacts. Studies in History and Philosophy of Science Part a., 37(1), 137–151.
Lenhard, J. (2007). Computer simulation: The cooperation between experimenting and modeling. Philosophy of Science., 74(2), 176–194.
Lenhard, J. (2019). Calculated Surprises: A Philosophy of Computer Simulation. Oxford University Press.
Morgan, M. S., Morrison, M., & Skinner, Q. (Eds.). (1999). Models as mediators: Perspectives on natural and social science. (Vol. 52). Cambridge University Press.
Morrison, M. (2015). Reconstructing reality. Oxford University Press.
Parker, W. S. (2003). Computer modeling in climate science: Experiment, explanation, pluralism (Doctoral dissertation, University of Pittsburgh).
Parker, W. S. (2009). Does matter really matter? Computer simulations, experiments, and materiality. Synthese, 169(3), 483–496.
Pincock, C. (2011). Mathematics and scientific representation. Oxford University Press.
Resch, M. M., Kaminski, A., & Gehring, P. (Eds.). (2017). The science and art of simulation I: Exploring-understanding-knowing. Springer.
Rohrlich, F. (1990). Computer simulation in the physical sciences. In PSA: Proceedings of the biennial meeting of the philosophy of science association (Vol. 1990, no. 2, pp. 507–518). Philosophy of Science Association.
Simon, H. A. (1996). The Sciences of the Artificial. MIT Press Books.
Symons, J. (2010). The individuality of artifacts and organisms. In History and philosophy of the life sciences (pp. 233–246).
Symons, J., & Alvarado, R. (2016). Can we trust big data? Applying philosophy of science to software. Big Data & Society, 3(2), 1–17.
Symons, J., & Alvarado, R. (2019). Epistemic entitlements and the practice of computer simulation. Minds and Machines, 29(1), 37–60.
Symons, J., & Boschetti, F. (2013). How computational models predict the behavior of complex systems. Foundations of Science, 18(4), 809–821.
Weisberg, M. (2012). Simulation and similarity: Using models to understand the world. Oxford University Press.
Weisberg, M. (2019). Abstraction and Representational Capacity in Computational Structures. The Scientific Imagination, 210.
Winsberg, E. (2010). Science in the age of computer simulation. University of Chicago Press.
Author information
Authors and Affiliations
Corresponding author
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
About this article
Cite this article
Alvarado, R. Computer Simulations as Scientific Instruments. Found Sci 27, 1183–1205 (2022). https://doi.org/10.1007/s10699-021-09812-2
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10699-021-09812-2