Exemplar-Driven Realism

In a recent analysis, Saatsi has argued that the realism debate is in the grip of what he calls ‘recipe realism', where such a recipe is ‘... capable of distilling the trustworthy aspects of a theory, applicable to any good, predictively successful mature theory.' (2016 p.

Thus, structural realists express their epistemic commitments in such general terms: ‘all that we know is structure'. And in that spirit, the debate then focuses on the letter of how that notion of structure may be metaphysically characterised and represented. Some, such as Worrall (1989), have tended to highlight the relevant equations and deploy Ramsey sentences as representational devices, whereas ‘on- tic' structural realists (Ladyman and Ross 2007; French 2014) have emphasised symmetries and the associated group-theoretic structure and have used the semantic approach to represent these features of theories. But Saatsi argues that the under­lying approach is the same: to give a general recipe that can be applied to any historical episode or any new scientific development (that meets the realists' general criteria of empirical success, maturity etc.).

Thus, the aim of recipe realism is ‘...to capture theories' epistemic commitments across a wide range of disciplines and different areas of scientific theorising in unified terms, offering recipes or algorithms that are first motivated by particular considerations and case-studies, and then optimistically projected well beyond those to the rest of science.' (Saatsi 2016, p.

Furthermore, Saatsi continues, the fact that we have so many different realist recipes should give us pause; perhaps, instead of thinking of them as competitors, we should regard them from a pluralistic perspective as capturing the different possible ways that a theory can ‘get the world right'—i.e. get it right in its structural aspects, in its causal aspects and so on. But of course, the fact that each recipe fits certain cases gives no reason to think it can be ‘projected' even within that par­ticular science, much less across disciplines, nor that the other recipes are somehow ruled out across the board.

And the very abstract nature of such recipes means they actually say little about precisely how the theories concerned latch onto reality. Again, in the case of structural realism, to give content to the recipe the nature of the structure that is appealed to must be spelled out but this has proven contentious, or so it is claimed. Here we have the contrasting examples of Ramsey sentences and the semantic approach, and it is argued that ‘[i]t is entirely unclear why these recipes should in general be at all good for discerning something that will be carried over in various theory changes, something that furthermore genuinely accounts for the past theo­ries' empirical success.' (ibid., p. xx) Even giving illustrative exemplars such as the famous Fresnel case or phlogiston, is of little help, since they may pull in different directions, thereby watering down the content of the recipe; after all, in what sense can ‘the structure' in the case of light and phlogiston be the same?

Nevertheless, Saatsi agrees that focussing on such exemplars is the way to go— except he recommends giving up on the recipe entirely: ‘Recipe realists are right in leaning heavily on exemplars in explicating their realist commitments, but they go wrong in trying to generate a general recipe that captures the gist of those exem­plars.' (ibid., p.

Now, there is much that is positive about this suggested reorientation of the realism debate. And indeed, despite being held up as an example of ‘recipe realism', there seems little to prevent structural realism from being articulated within the framework of ‘exemplar realism'. Although the epistemic form of structural realism was indeed originally presented as a general and perhaps abstract response to the Pessimistic Meta-Induction (Worrall op. cit.), the ontic variant allies that response to an attempt to accommodate the specific metaphysical implications of quantum mechanics (Ladyman 1998). And as has been noted (French 2006), that alliance may in fact come apart. Consider: towards the end of his classic paper, Worrall speculated that his form of structural realism might be extended to quantum phy­sics. Now, although attempts have been made to articulate an appropriate sense of continuity between classical and quantum mechanics (see Saunders 1993, French 2014 pp. 15-20), it might be argued that such attempts can only be deemed to be successful to the extent that we accept a certain ‘plasticity’ in the relevant structures (so that the Poisson brackets of classical mechanics can be deemed to be appro­priately related to the Moyal brackets on the quantum side for example; see French ibid.). If one were to conclude that the bridge between the two is just too tenuous one might then be inclined to conclude that either one should acknowledge that the structures one should be realist about are different in the classical and quantum cases, with no relevant continuity between them (and hence convergent realism is in trouble, as Laudan famously suggested), or that if one has grounds to be a structural realist when it comes to quantum physics, one may have no such grounds in the classical context, where one should be an entity realist perhaps.

And of course, setting aside the issue of responding to the Pessimistic Meta-Induction, the relevant grounds for shifting to structures will vary both within a particular discipline, such as physics, and between disciplines, such as physics and biology. Thus when it comes to quantum mechanics, part of the original motivation for ontic structural realism concerned the perceived metaphysical underdetermination between the views of particles as individuals and as non-individuals, both of which are supported by the relevant quantum statistics (see Ladyman 1998; Ladyman and Ross 2007; French and Ladyman 2011; French 2014). In response to van Fraassen’s conclusion that such underdetermination undermined the realist’s position, it was argued that it could be effectively dissolved by giving up on the underlying object-oriented metaphysics and claiming that, to repeat the slogan, ‘all that there is, is structure’ (Ladyman 1998). This is less a ‘recipe’ than a metaphysical commitment that of course then needs to be cashed out.

However, that cashing out of what is meant by ‘structure’ should not be understood in terms of Ramsey sentences or set-theoretic structures or category theory or whatever. As French (2014, Chap. 5) emphasises, these are the devices that we use as philosophers of science to represent, for our own purposes and aims, theories, data models, programmes and, yes, empirical and theoretical structures. But our realist commitments should not be to these devices in and of themselves. In deploying them we are not so much giving content to the relevant ‘recipe’ as using a meta-level tool. Thus it should come as no surprise that different philosophers of science, with their different meta-level commitments, should use different sets of such tools. Worrall, for example, is famously antagonistic towards the semantic approach and thus prefers the syntactic formulation of Ramsey sentences to capture the structural commitments manifested at the object level of the theories themselves in the form of the relevant equations.

frameworks (Bain 2013; Landry 2007), although these have been criticised for failing to appropriately represent the relevant structural features (Lal and Teh 2015; Lam and Wuthrich 2015). And, of course, the set theoretically based semantic approach has long been advocated as an appropriate means of capturing the inter-theoretic commonalities that are claimed to form the basis of the structural realist's response to the Pessimistic Meta-Induction (Ladyman 1998; French 2014, Chap. 5)

Furthermore, that cashing out at the ‘object level' of the science itself will be specific to the relevant theoretical context. Thus in the context of the theory of light and, subsequently, electromagnetism, the relevant structure is presented by Worrall in terms of the equations of first, Fresnel, and then Maxwell (and beyond), taken to be interpreted of course. And although we find such features as Galilean invariance in the classical context, it is in the quantum physics that symmetries really come to prominence, beginning with the permutation invariance that lies at the heart of the quantum statistics underpinning the above metaphysical underdetermination. And just as the laws are presented mathematically via the appropriate differential equations, for example, these symmetries are presented via the mathematics of group theory. Hence in this case the ‘structure' is cashed out in terms of the relevant laws plus symmetry principles (see French 2014),^[72] where these are then clothed in an appropriate metaphysics.

Shifting to quantum field theory, we no longer have the original motivation in the form of the above metaphysical underdetermination (although we do have another kind of underdetermination in the form of fields-as-substantival versus fields-as-instantiated properties) but we still understand the structure in play through a combination of symmetries and laws, with the Poincare symmetry of relativistic space-time playing a particularly significant role (French and Ladyman 2003; French 2014).

So, in one sense, we don't have the same recipe cashed out in each case, since the motivation for structural realism that is presented in the context of quantum mechanics is not present in that of high-energy physics, where the motivation has more to do with the way that fundamental properties such as spin and charge effectively ‘drop out' of the relevant symmetries. But of course, in another sense it can be alleged that we do, insofar as it is the relevant symmetries that are focussed on in each case, as presented in the theoretical contexts by the appropriate groups.

However, one can insist in response that this is entirely driven by the relevant context not by some adherence to a particular realist recipe: it is because of the framework provided by quantum field theory that we find Poincare symmetry also playing a fundamental role in the context of the Standard Model and it is because of the role of gauge symmetries more generally that we find the notion of structure cashed out in this context in this manner as well. In other words, what appears to be the same recipe applied again and again is in fact due to the features of the relevant physical theories.

But then of course we should not expect these same features to be exemplified either by other theories within physics or by the theories of other disciplines. So, no one of course would claim that when it comes to theories of light and phlogiston the structure is the same. Indeed, insofar as the latter example might be seen as falling under ‘chemistry’ (and here we might need to be sensitive to disciplinary bound­aries), we would clearly not expect to encounter the same equations or laws much less any symmetries (see French 2014 Sect. 12.2). Likewise when it comes to biology, where we not only have no symmetries but no laws either, except perhaps for natural selection. Nevertheless, although we clearly no longer have the moti­vation for shifting away from objects that was articulated in the quantum context, the kinds of concerns with the nature and role of the notion of ‘organism’ and biological object more generally that have been articulated by Clarke, Dupre and others (see, for example, the papers in Guay and Pradeu 2016) have been taken to power a similar shift from understanding biological entities in object oriented ways to conceiving of them in terms of certain kinds of biological structures and pro­cesses (French 2014, Chap. 12; French 2016). Again it might be emphasised that it is reflection on the science itself rather than sticking to a particular realist recipe that is driving these moves.

All of which amounts to saying that in certain respects structural realism is already exemplar-driven and there seems to be no inherent barrier to rendering it explicitly so. Thus, from this perspective, the structuralist would acknowledge that, at the very least, the motivations and reasons for this shift will vary from context to context and discipline to discipline and indeed that in some cases there simply will be no such grounds. In other words, whether structural realism is the appropriate stance to adopt would have to be tested on a case by case basis.

However, if our realism is going to be exemplar based then there is even greater need to be clear on what it is we are going to be realist about. Consider: the structural realist has long pointed out that underlying the ‘recipe’ of standard realism is a certain kind of ‘object orientation’. In effect this smuggles in an implicit metaphysics so that when the standard realist declaims ‘I am a realist about elec­trons’ and is then pressed on what these electrons are, she can then say ‘they are objects, like tables and chairs, albeit subject to the laws of quantum physics which make them behave in weird ways...’ As far as the structural realist is concerned, the object oriented standard realist gets away with a lot by means of this manoeuvre, since she never seems to face the equivalent to ‘what is this ‘structure’ of which you speak?’; that is, she never seems to have to answer ‘what are these objects that electrons are supposed to be?' In other words, the recipe masks the underlying metaphysics.

If that mask is then stripped away and we ground our realist stance in distinct exemplars, then we cannot get away with keeping the metaphysics implicit—it must be stated explicitly in each case. The alternative is to adhere to an entirely epistemic form of realism (or what Magnus calls ‘shallow' realism; Magnus 2012) which would amount to pointing, if pressed, to the relevant features of the theory, as expressed in its equations or models or whatever, and insisting ‘I am a realist about that!'. But as a response to the demand to say how the world is according to the theory, that hardly seems adequate. Hence we need to appeal to some appropriate metaphysics in each case. The question then is how to avail ourselves of that metaphysics.

3