- Broadening the Perspective:Epistemic, Social, and Historical Aspects of Scientific Modelling
The recognition that models and simulations play a central role in the epistemology of science is about fifteen years old. Although models had long been discussed as possible foundational units in the logical analysis of scientific knowledge, the philosophical study of modelling as a distinct epistemic practice really got going in the wake of the Models as Mediators anthology edited by Margaret Morrison and Mary Morgan (1999). In spite of the broad agreement that in fact much of science is model-based, however, there is still little agreement on pretty much anything else. What are models? Are they representations or fictions, abstract entities or concrete artifacts? Which functions do they play? Can they explain, provide confirmation to hypotheses, or are they mere heuristic devices? If they have independent epistemic power, where does it come from? Moreover, arguments in favor or against alternative positions are often drawn from case studies of particular modelling or simulation practices. Finally, those philosophical accounts that aim at generality often tend to be “deflationary” in spirit since, it is claimed, various kinds of models do not share enough in common to warrant a substantive and yet unified view.
How then is progress in understanding the epistemology and ontology of models to be achieved? Several routes are possible. One route is to devote further philosophical effort into the development of a general account of scientific models, their ontology and epistemology. Another is to examine more, and more diverse, cases of scientific models and aim for better taxonomies and informative generalizations. Yet another is to [End Page 381] broaden the range of theoretical resources so as to include cognitive, social and historical perspectives on models and modelling.1 In this special issue, comprising a selection of papers presented at the conference “Models and Simulations 5” (Helsinki, 2012), all three routes are explored. This introduction is an attempt to highlight the ways in which the various perspectives can work productively together.
II. The Epistemic Role of Models
Models are means of surrogate reasoning: we learn about the systems modelled by examining and manipulating an epistemic artifact purposefully constructed to “stand in” for the system. If models can be thought of representing their target, the way in which the model functions as a stand in for its target can hardly be understood independently of its use and purpose. More generally, pragmatic factors are indispensable in defining what the model is a model of as well as its accuracy or success (e.g. Suárez 2004). Recent philosophical interest in modelling was mostly sparked by the realization that models are in many ways autonomous with respect to theory, the unit of knowledge long held to be foundational for science. Models are neither straightforwardly derived from theory nor simply abstracted from observation and experiment.
In their representational capacity models can be thought of as analogous to maps in that they have both a purely iconic or mimetic quality as well as a conventional element. Lehnard (this issue) explores features of computational models brought forth by the map analogy in the case of nanotechnology. First, Lehnard argues that the results of computational modelling do not produce a global representation, but are conceived of as parts of “virtual atlases” made out of local, purpose-dependent maps, which do not fit neatly together. The success of maps does not rest on how accurately they represent their target, but on whether they provide correct predictions (or retrodictions) and suggest effective intervention. Second, maps and computational models are products of negotiation: in the historical past, map making involved negotiation between, say, the cartographer and his/her informants, computational modelling results from a process of negotiation between theory, simulation, and experiment (on the interaction between theory, models and experiments see also Barwich, this issue). The third feature is ‘locality’: the simulation model is not straightforwardly derived from general laws nor does its success [End Page 382] in one situation and under certain conditions imply success in others; in this sense it is local rather than general.
The epistemic justification of different kinds of models need not be the...