1 | THE RISE AND FALL OF DEDUCTION
R, is We can therefore supplement PS with an answer, at least in an explanatory context That
is, R is a Waltonian prop but constrained by modal requirements. So models are not structures, they are fictions (in Frigg's ontological sense), but they manage to represent through the modal explanatory constraints we impose on them, (potentially those modal constraints themselves
Set-theoretic model of target system/data Set-theoretic model of model Target system/data R R R (asymmetric) Partial- isomorphism (symmetric) Scientific model
scientific representation is determined by a particular explanandum. One might worry that this is to conflate two things, after all explanation is only one motivation of a scientist, there are many others: to teach, to represent the true ontology of the world, to predict. This is true, but remember that in this framework prediction is also fundamentally about providing modal information, just future directed modal information. Prediction and explanation in the real world are different but in side model worlds they are the same, or at least on a par. Therefore, if a scientist represents in a way that is totally untethered to either prediction or explanation then modal constraints will be irrelevant. My contention here though is that these cases are in the minority. If one wishes to contest this assertion then so be it, there is nothing much at stake, my claims still apply to any model that has explanation or prediction as a desired use, and the reader can take it as shorthand that from now on the word "model" applies only to this subset (and decide for themselves how large a subset of actual scientific models this is). Representation, in this explanatory context, becomes a secondary quantity confined by the needs of explanation/prediction. These modal constraints are given a unitary framework, while the representational apparatus, in reductive terms, can be given a pragmatic heterogeneous treatment.
166 A natural instinct is to ask what it is that allows a model to capture modal structure. Fine's response to this (1993) is that explanatory requests must be judged by the paradigms they come out of. Fictional models are not epistemically disconnected from us in the way mathematical abstracta as traditional conceived seem to be. For Fine the demand for further information is not legitimate, models will work in different ways, in each case a scientist will have a reason why that model is explanatory and demanding a philosophical one-size-fits-all philosophical reconstruction is unreasonable without justifying that need.
I have some sympathy with this view, to explain is to provide modal information. If our aim is to understand how science works then requests for ultimate justification may be misplaced. Just as a causal explanation cannot be called inadequate because a researcher cannot supply a causal chain that reaches all the way back to the Big Bang, or all the way down to the ultimate level of reality, so too it is perhaps asking too much to demand always a further step behind 'the Wizard's curtain'. That said, I think one can say a little more, and that will be the subject of chapter 8.
167 Morrison has also stressed the highly context dependent nature of explanatory fictions (1999). 168 It is not clear if this is Bokulich's view, sometimes modal structure is stressed other times
dynamical structure.
169 Consider the various different models of the nucleus and their different explanatory targets. By producing such an abstract steam engine Carnot was able to see the reversibility and symmetry in the stages of how the heat engine works, something which would have been obscured by the particular details of a real working heat engine. Irrelevances, just as in the allometric scaling law case, are not harmless, often they obscure fundamental symmetries and parallels between otherwise very diverse phenomena.
7 | PERSPECTIVISM
170 I am not arguing that other strategies aimed at realism are not viable or compatible with the notion of explanation mooted here, only that as a general holistic framework a form of surrealism should be our attitude which is supplemented on a case by case basis with other realist strategies that do not invoke explanatory success.
171 Of course one could argue that nihil models are not explanatory at all. The problem with this is that the dividing line between explanatory and non-explanatory models is entirely historically contingent, the known fictions of semi-classical models make them non-explanations, but theories in which elements subsequently turn out to be fictions cease to be explanatory. I believe Newton's theory of gravity is explanatory and what is an explanation is not a judgement that can only be made at the end of science.
172 Van Fraassen 1980. 173 French and Ladyman 2003. 174 Bueno 1999.
175 This view is very much in the spirit of many of Woodward's views on explanation, my aim is to add formalism, albeit in small doses, to the mixing of ontic and epistemic qualities in his account.
176 The perspectivism argued for here is a very distant cousin of the historical perspectivism advocated by Nietzsche, although my aim aim is to throw the bath water out of that account and leave the baby relatively unscathed!
177 For instance, Woodward (2003) has argued that evolution has shaped our brains to work in certain ways, one of which is to manipulate tools, and this predisposition to think in terms of manipulation has formed part of our conceptualisation of causality. Manipulationism is not ahistorical: history takes a seed, our innate ability to manipulate our own bodies, then simple tools, and builds a logic of causal explanation out of them, and then theoretical non-causal explanation. The task is not to deny history, to pretend that cave men and women were reasoning in the way described, or even early scientists were, rather it is to identify the seeds of
178 I am not arguing here that this is the case, only that such suggestions are entertained seriously to illustrate the potentially huge disconnect between how the world actually is and how we see it. 179 Of course some interpretations, such as that suggested by Price are so radical in their
interpretation of time they undermine the possibility of the scientific method that lead them to their conclusions about it. In Price's scheme, for instance, future events near the end of the universe might well dictate current particle trajectories. If this is the case it is hard to see how experimental science could have any validity, and hence how we could have any evidence for the atemporal physics Price places so much emphasis on.
180 Giere 2006. 181 Brown 2009. 182 Feyerabend 1999. 183 Dewey 2007.
184 Unsurprisingly. I claim that this perspectivism is best served by thinking in terms of modal topology. When we shift perspectives we have to preserve the 'topology' of the modal connections. That is, if we have a parameter set and a list of connections between those sets, then move to a new framework and keep some, or all, of those parameters we must also preserve the modal connections, at least for the domain in which our theory appears explanatory. Why call it 'topology'? To convey the idea that not all of the modal structure has to be carried over, but some essential connections must be. We can add intermediate variables/processes, things which may for instance allow a more fine grained teasing out of modal intricacies, but we must also keep in global terms the modal connections of the previous model. When we move from the ideal gas treatment to a van der Waals model we have many of the same input/output variables to connect modally in the models and each must agree on which variables are connected to which others, even though the van der Waals case adds new variables, new axes of modal variation as it were, extra 'nodes' that can be linked in a modal map. The word topology is use loosely to convey the flexibility in how the modal structure of an explanation and a successor explanation can each display modal information differently but keep in common some essential connections that represent what-depends-upon-what. To use the analogy of Beck's underground map mention in chapter 5, new stations can be added or subtracted and the shape of the lines changed but the stations cannot be moved from one line to another if an explanation has really got some essential modal connections correct.
185 Rather like the radical sceptic, the preferences revealed by practice by such thinkers is to treat tables as if they exist for the purpose of holding up their tea cups.
186 See Morrison 1999 for more details.
187 Those modal connections are in models of course and then can be tested to see if they capture modal information in the world, to see if the perspective adopted is a modally useful one. 188 For instance we can ask and answer w-questions about the transition of variables form
upstream to downstream which might produce different spectral outputs due to different heating profiles. Although the jump conditions must be satisfied there are potentially many different ways of connecting them and when comparing to an actual observed shock region only one of these may explain the observations. The key point is that the global answers to w- questions, the solutions of the jump conditions, remain the same, but we add more intermediate modal nodes with a more detail model to answer more w-questions.
189 The use of interpellation is intended loosely, it is not meant to imply that scientific perspectives are like notions such as national identity. The scientific perspectives are born out of the need to explain. This explanatory constraint is fundamental. We have a kind of modally objective interpellation, in the sense that our basic epistemic perspectives are added to by abstract conceptualisations that allow new modal connections to be discovered.
190 Knuuttila 2005
191 If one wished it might be possible to argue that a Kuhnian paradigm shift should be re- conceived as not a radical discontinuous change in theory but in perspective, and once new perspectives are developed scientists in that field are interpellated into seeing, or noticing, that perspective in other systems. For instance, the way the Ising model is useful way beyond its intended target system. Perspectives are a looser notion than paradigms, the main function of a perspective is to allow a variable parametrisation such that modal connections can be captured. 192 In Giere's terms this account might be dubbed modal surrealism.
193 Of course Poincarean conventionalism about the shape of space is possible. We actually do have a universal gravitational field and a flat space-time, but this does not change the point which is the potential for which parts of a theory we take ontologically seriously to shift with perspective. 194 It is perhaps a classic example of where a 'trading zone' needs to be set up between
communities to agree the properties that are common to different uses of the same signifier for different signified.
195 One may argue that in principle the molecular genetic framework can answer all of the w- questions the classical framework can, that is not the contention here, but if if that is true answering w-questions in principle is not to answer them at all! Remember that our focus is explanation, explanation through model information: if one cannot actually say what-depends-
details would obscure the modal connections revealed at a coarse-grained level. It is clearly part of the conception of modal topology that detailed molecular explanations of particular elements of those explanations cannot be in conflict. and may allow a wider set of w-questions to be answered. Such an observation is trivially reductionist, the explanations only reduce, if at all, within a historical sequence, the more abstract model being used to identify relevant partitions to connect modally in the first place. Those partitions may be translatable into lower level terms but they could never have been identified from that lower level perspective. Often tractability is presented as some merely pragmatic matter, what I am arguing for is the claim that tractability has a much more fundamental role to play than expressing the calculational limits of our latest supercomputer.
196 Of course in what sense do we get things modally correct when we use classical genes? We can breed peas and manipulate them, and get outcomes of those manipulations but not by manipulating classical genes directly, not experimentally, after all they don't exist! But how do we tell the difference between classical and molecular? It is very difficult. Let us say both are explanatory, then we can can m-manipulate classical genes, and the t-manipulation involves manipulating molecular genes (if such a thing exists), the phantom manipulations supervene upon real manipulations. We can see how the older explanation can still be explanatory, it managed to capture some modal information, but our current theory could be just the same, and hence we should adopt a certain level of surrealism. If it looks as if classical genes exist, so too it may just be that it looks as if molecular genes exist, but this is part of the perspectivism: classical gene explanation doesn’t stop being an explanation because we have molecular genes. For instance, evolutionary biologists explain lots of things invoking the classical gene, because when it comes to modally linking variables from the conceptual parametrisations of evolutionary theory the world looks as if classical genes exist, that is enough for those explanations to be explanatory in an ontic sense.