[for a brief explanation of this ongoing series, as well as a full table of contents, go here]
Philosophy of science: forms of realism and antirealism
I’m a philosopher of science, and therefore better acquainted with that subfield than with anything else in philosophy. And it is clear to me that philosophy of science has also made quite a bit of progress from its inception (the approximate time of which may reasonably be pegged onto the famous debate on the nature of induction between John Stuart Mill and William Whewell, but which could be traced further back at the least to Bacon). In this section I will go over a specific example (among many, really) of progress in philosophy of science, the debate between realists and antirealists about scientific theories and unobservable entities (interested readers should also see the many useful references listed in the pertinent SEP entries: Monton and Mohler 2008; Ladyman 2009; Chakravartty 2011).
To begin with, let us be clear about what sort of debate this is: it concerns what our best epistemic attitude should be toward scientific theories, as well as toward the “unobservable” entities posed by most of these theories, such as electrons, strings and so forth. The broad contrast is between realist and antirealist positions, though each one comes in a variety of flavors and nuances. Within the context of this discussion, “observable” means by way of normal human sensorial experience (as opposed to scientifically observable): stars and dinosaur bones are observable; electrons and galaxies are not, in this sense. While the distinction so drawn is obviously arbitrary, since it is defined by the sort of human sensorial access to the world, that is precisely the point: the realist-antirealist discussion is one that concerns human epistemic limits within the context of science, as well as the degree to which our ontology should be justified by our epistemology.
Scientific realism comes with a number of commitments: epistemically, it takes scientific theories to yield knowledge about the world — as opposed to a more limited instrumentalist interpretation of those theories, which at most grants that we can have knowledge of observables, but not of unobservables. Metaphysically, for a scientific realist the world exists independently of human minds, a stance which is of course open to the Kantian objection that we do not have access to the world-in-itself, only to decidedly not mind-independent experiences. Semantically, scientific theories are to be taken at face value for the realist, as statements about the way the world is — as opposed to, again, instrumentally (i.e., they just work, in a pragmatic sense). Of course realists do typically admit that most scientific theories are, strictly speaking, false, which is why they often appeal to the notion of “truth-likeness,” rather than truth (see Chapter 4). As a result of this, realists tend to be fallibilists about scientific knowledge but nonetheless maintain that science makes progress by getting incrementally — though not necessarily linearly, or even always monotonically — closer to a true description of the world.
There are at the least three major varieties of realism, which are committed to stronger or weaker versions of the above. Entity realism holds that we have good reasons to be realists about any entity described by a scientific theory of which we have successful causal knowledge (e.g., black holes in cosmology); explanationism is the idea that we should be committed to realism with respect to the “best,” most indispensable, parts of scientific theories (e.g., quarks in fundamental physics); and structural realism is the position that we should be realist not about the entities invoked by scientific theories, but about the (mathematical) structure of those theories (e.g., treating the equations of Newtonian mechanics as a limit case of those from Special Relativity).
There are a number of arguments in favor of and against scientific realism, and the field has made progress precisely insofar these arguments have been presented, criticized, modified, further criticized, refined, and so forth, the by now familiar exploration by move and counter-move of philosophy’s conceptual spaces. Briefly, there are three major reasons typically deployed in favor of scientific realism, the best known of which is Putnam’s (1975, 73) “no miracles” argument: “[realism] is the only philosophy that doesn’t make the success of science a miracle.” To this, van Fraassen (1980) — a major critic of realism — responded that a different way of thinking about the issue is to consider good scientific theories as the conceptual equivalents of successful living organisms: adapted to their environment and therefore effective. That, in turn, raises the issue of what, exactly, accounts for the adaptness of a scientific theory, not an easy issue to address, given that we still lack a satisfying explanation of cultural evolution couched in selective terms (though plenty of people have been working on it). Another objection raised against the no miracles argument is constructed on a version of the base rate fallacy: we just don’t know what the baseline expectation for successful scientific theories is (i.e., we do not have an independent control against which to make assertions of success or failure), so on what grounds are we surprised by the effectiveness of science? I confess that I find this objection formally intriguing, but ultimately bizarre, unless one wishes to deny the truly astounding degree to which science has made progress compared to any other way that human beings have devised to learn things about the natural world.
The second standard argument in favor of realism is corroboration, the idea that if different means of empirical detection, or different theoretical approaches, converge on the same conclusions then we have good reasons to be realists about those conclusions, a meta-theoretical version of Whewell’s (1847) famous consilience of induction. For instance, Ladyman and Ross (2009) have argued that there are several theoretical invariants in fundamental physics, regardless of which of a number of currently competing theories about the basic structure of reality turns out to be true. It seems sensible, therefore, to suggest that those theoretical invariants will survive independently of which of the competing theories will eventually be settled on by fundamental physicists (this is a type of structural realism).
Finally, we have selective optimism, the suggestion that we can retain entities or theoretical structures from previous theories, if they keep playing a successful role in new theories, which in turn gives us reasons to be realists about those entities or structures. In the case of entities, Ladyman (2009) suggests that this nicely dovetails with the kind of theories of reference proposed by Putnam (1975) and Kripke (1980), which maintain that one may retain successful reference to certain entities (e.g., electrons) even in spite of substantial background theoretical changes, leading to a certain stability of epistemic commitments even in the face of theory change.
On the other side of the debate, there are well articulated (and just as carefully criticized) arguments against scientific realism. Perhaps the best known is the problem posed by the underdetermination of theory by data, which is closely related to the Duhem-Quine theses (Chapter 2). The idea is that there are always, in principle, many different theories that are equally compatible with the data, or — equivalently — that the available data is not enough to determine the truth of a theory and the falsity of closely related theoretical variants. The problem with this objection is that it is actually difficult to find historical examples of rampant underdetermination, or of an underdetermination that was not soon resolved. Then again, perhaps the best current example of massive underdetermination of theory by the data is string/M-theory in fundamental physics (Woit 2006; Smolin 2007; Baggott 2013), with its “landscape” of 10^500 (give or take) possible versions and essentially no empirical evidence to even tell us whether the broad outline of the story is correct.
A second standard objection to scientific realism is rooted in skepticism concerning the idea of the above mentioned inference to the best explanation (IBE), which can be articulated on two levels. First, there is criticism of the very concept of IBE, based on the difficulty to make explicit what criteria are to be deployed in order to determine that a given explanation is, indeed, “the best.” Should we use simplicity? Elegance? A combination of these, and what else? The matter is actually not at all trivial, though my sense is that most philosophers do think that IBE is a defensible type of inductive inference. Second, van Fraassen (1989) noted that at any given point in time in the history of science we may just have a “best of a bad lot” situation, so that even if we had a principled way to make an inference to the “best” (available) explanation, we would have no reassurance that such explanation was even in the ballpark of the true one.
Then we have the problem (for realists) posed by the so-called pessimistic meta-induction: Laudan (1981b) pointed out that the history of science provides us with a number of examples of theories that were thought to be correct and were eventually rejected. Thus, by standard inductive generalization one would be justified in concluding that current theories will also, eventually, turn out to be false. If there is no principled way to address this, then an epistemically modest attitude toward scientific theorizing — such as antirealism — seems warranted. The pessimistic meta-induction can be countered by deploying the notion of truth-likeness, the idea that we are getting incrementally closer to the truth in a way that can be argued without falling into question begging, and which I have already discussed in Chapter 4. Needless to say, antirealists have also criticized the very notion of truth-likeness.
So much for the basic version of scientific realism and its main flavors. Let me turn now to a symmetric treatment of the major alternative proposal, van Fraassen’s constructive empiricism. The opening salvo for a renewed attack on scientific realism was his The Scientific Image (1980, 12), at the onset of which he unequivocally threw the gauntlet down: “Science aims to give us theories which are empirically adequate; and acceptance of a theory involves as belief only that it is empirically adequate.” It is this book that is usually credited with the rehabilitation of the very idea of scientific antirealism, after the demise of logical positivism (Stadler 2012).
Constructive empiricism has much in common with logical positivism, particularly the commitment to purge what are perceived as unnecessary metaphysical burdens from our view of science; however, van Fraassen’s position — unlike logical positivism — does not rely on the (in)famous verification principle, nor does it shy away from the idea that scientific discourse is necessarily theory-laden. The crucial notion is that of empirical adequacy, which van Fraassen characterizes in this manner: “a theory is empirically adequate exactly if what it says about the observable things and events in the world is true — exactly if it ‘saves the phenomena’” (1980, 12). But constructive empiricism advances a number of additional ancillary notions, a particularly important one being the distinction between observables and unobservables, where the line separating the two categories of facts or entities is understood as relative to human beings qua epistemic agents.
As in the case of scientific realism, over time several arguments have been put forth in favor of constructive empiricism, and naturally many such arguments mirror the ones we have seen above about realism. We do not therefore need to revisit the idea of underdetermination of theory by data, or the pessimistic meta-induction. Beyond those, a major point highlighted by van Fraassen is that his preferred criterion of empirical adequacy is epistemically more modest — so to speak — than claims of truth, while still making sense of science as an epistemically successful activity. Another major difference between constructive empiricism and scientific realism is the way these two frameworks approach the proper relation between theory and experiment: realists think that experiments teach us about observables as well as unobservables, while constructive empiricists think that experiments only teach us about the former.
Pragmatism is big among constructive empiricists. They, for instance, follow a pragmatic approach to theory choice: van Fraassen rather astutely observes that scientists themselves often deploy criteria for theory selection that are not strictly epistemic, but — in fact — pragmatic, for instance, when a theory is preferred over another on grounds of simplicity, elegance, etc. He also advances a pragmatic view of explanation, providing a long list of past scientific theories that had explanatory power and yet turned out not to be true (this is conceptually distinct from the point about the pessimistic meta-induction): e.g., Newton’s theory of gravitation explained the movements of the planets, and Bohr’s theory explained the spectrum emitted by hydrogen atoms. It seems uncontroversial that scientific explanations do include a pragmatic component, but — adds the constructive empiricist — how can the scientific realist insist that they are good explanations because they are true statements about the world, given examples such as these and a number of others that can easily be gleaned from the history of science?
Indeed, van Fraassen’s pragmatism even extends to the very idea of laws of nature, which he regards as unnecessary metaphysical commitments, in explanatory terms, to which nonetheless the scientific realist seems wedded. Constructive empiricists point out that natural laws are actually only true on a ceteris paribus clause (a point famously also made by Nancy Cartwright: 1983), that is, they depend crucially on counterfactuals that are — strictly speaking — empirically inaccessible. Both the emphasis on empirical adequacy and the frontal assault on the sacred cow of natural laws is part and parcel of one of the major points in favor of constructive empiricism, according to its supporters: a philosophical program that puts a stop to “inflationary metaphysics.” As van Fraassen (1980, 69) clearly explains: “the assertion of empirical adequacy is a great deal weaker than the assertion of truth, and the restraint to acceptance delivers us from metaphysics.” The idea is that the constructive empiricist can drop concepts such as those of laws of nature, natural kinds and the like, without any apparent loss in explanatory adequacy.
As in the case of scientific realism, there are a series of moves in conceptual space that various authors have made to dispute constructive empiricism — with the full ensemble of these moves and countermoves representing another instance of what I call progress in philosophy. We have already encountered some of these maneuvers, beginning with Hilary Putnam’s no-miracles argument, and with the scientific realist’s invocation of inference to the best explanation.
A less successful charge against constructive empiricism was led by Paul Churchland (1985), and it fundamentally hinged on the arbitrariness of the observable / unobservable divide drawn by van Fraassen. While prima facie reasonable, Churchland’s objection actually misses van Fraassen’s point: as explained above, the latter was not attempting to draw a hard metaphysical line between observable and unobservable, but simply making the more mundane empiricist observation that scientific knowledge is human knowledge, and therefore bounded by the limits of humans qua epistemic agents, including the by no means arcane fact that we can observe certain things and not others.
Be that as it may, a recent survey of philosophers’ positions (Bourget and Chalmers 2013) clearly gives realism the upper hand over anti-realism (75% vs 12%, give or take — see next Chapter for more details), though the survey itself does not discriminate among types of realism and anti-realism. One of the factors contributing to this disparity may be the next big leap in the exploration of the realism-antirealism conceptual space, to which I now turn: structural realism (Ladyman 2009), a conception first introduced by John Worrall in 1989, a few years after van Fraassen’s book.
Again recall that one of the most convincing grounds in favor of realism is Putnam’s no-miracles argument, and one of the strongest rebuttals by antirealists relies on some version of the pessimistic meta-induction, or at the least on the observation that major scientific theories of the past have been abandoned (something often referred to as “radical theory change”). It was therefore perhaps inevitable that the following major counter-move by realists is a novel type of challenge to the idea of radical theory change. Here is how Worrall (1989, 117) himself put it, in the case of a specific historical instance: “There was an important element of continuity in the shift from Fresnel to Maxwell — and this was much more than a simple question of carrying over the successful empirical content into the new theory. At the same time it was rather less than a carrying over of the full theoretical content or full theoretical mechanisms (even in ‘approximate’ form) … There was continuity or accumulation in the shift, but the continuity is one of form or structure, not of content.” That is, the commitment by the realist should be to the mathematical or structural content of scientific theories — what, allegedly, gets conserved across theory change — rather than to specific unobservable entities or descriptions of those entities.
From a historical perspective, there are indeed a number of apparent cases of successful structural transition between scientific theories, besides Fresnel => Maxwell. For instance, Saunders (1993) has argued that there are surprising structural similarities between Ptolemaic and Copernican astronomy, as counterintuitive as that may seem. More convincingly, perhaps, Bohr famously argued that classical models in physics are a limit case of quantum mechanical ones, and the same is true of the relationship between classical mechanics and Special Relativity.
It is important to distinguish between two major varieties of structural realism, epistemic and ontic, each representing — in the model of progress in philosophy that we are exploring — a distinct local peak in the conceptual mountain range that defines the realism-antirealism debate. In the case of epistemic structural realism, the idea is that one is justified in being a realist about the relations between unobservables, while remaining agnostic as to their ontological status. Intellectual forerunners of this approach date back at the least to the beginning of the 20th century (e.g., Poincaré), and can be read according to a Kantian key: science gives us access to the structure of the noumenal world, but not to its substance. More recently, Maxwell (1972) has defended a modern interpretation of epistemic structural realism deploying the concept of Ramsey sentences (Papineau 2010; Frigg and Votsis 2011) and turning it into a semantic theory. However, Ladyman (1998) has argued that this doesn’t actually constitute much of an improvement over standard structural realism, and that a stronger, metaphysical (rather than just epistemological) move is necessary.
Which brings us to the ontic approach. As Worrall phrased it (1989, 57): “[O]ur science comes closest to comprehending ‘the real,’ not in its account of ‘substances’ and their kinds, but in its account of the ‘forms’ which phenomena ‘imitate’ (for ‘forms’ read ‘theoretical structures,’ for ‘imitate,’ ‘are represented by’).” The focus, in other words, is on the structure of the relations among things, not on the things themselves — hence the title of Ladyman and Ross’(2009) book, Every Thing Must Go. Indeed, ontic structural realism can be seen as a “naturalized metaphysics” based on the most current accounts of the fundamental nature of the world provided by physical theories where, because of phenomena such as quantum non-locality and entanglement, the deeper one digs the less one finds anything like physical objects at the bottom of it all.
There are, predictably, a number of objections that have been moved to ontic structural realism, a fundamental one being that it endorses a strange metaphysics where there are relations without corresponding relata, which strikes many commentators as being this (visualize a very very short distance between your index finger and thumb) close to absurdity. I do not wish to get into the details of this sort of discussion here, though the various criticisms and counter-criticism of ontic structural realism represent a micro-cosmic version of my general thesis about movement in philosophical conceptual space. Ladyman (2009) lists and discusses seven different ways available to supporters of ontic structural realism to respond to the relations-relata problem, ranging from biting the bullet and going Platonic, so to speak, to the idea that individual objects do not have intrinsic natures after all, to a rejection of Humean-type supervenience, just to mention a few. Regardless, and as I’ve stated above, a major point in favor of ontic structural realism is that it is consistent with — indeed it is partly inspired by — the view of reality that originates from current fundamental physics. In particular, physicists nowadays think in terms of fields, rather than particles, since particles, in a quantum mechanical sense, don’t really exist as individual entities and are better thought of as field attributes of space-time points, which sounds a lot like there are no “things” or relata, just relations.
Irrespective of the debate about relations sans relata, there are several other objections to ontic structural realism: without going too deep (see review by Ladyman 2009), Chakravartty (2004) is among those that argue that sometimes structure too is lost during theory change, resulting in so-called Kuhn-losses (Post 1971, Chapter 4). Another issue is that ontic structural realism essentially turns metaphysics into epistemology (which may or may not be a good thing in and of itself, depending on one’s philosophical leanings), because it defers too much to the results of the natural sciences (especially, or, rather, pretty much exclusively, fundamental physics). And the proper relationship between metaphysics and epistemology is, naturally, a whole separate area for discussion (Chalmers et al. 2009; Ross et al. 2013). Moreover, some critics (e.g., Chakravartty 2003) argue that ontic structural realism cannot account for causality, which notoriously plays little or no role in fundamental physics, and yet is crucial in every other science. For supporters like Ladyman causality is a concept pragmatically deployed by the “special” sciences (i.e., everything but fundamental physics), yet not ontologically fundamental.
A more serious objection, I think, is that even modern physics still lacks a way to “recover” macroscopic individuality from quantum non-locality (after all you and I are individuals in a rather strong sense of the term, unlike tangled electrons). Without this account, both fundamental physics and ontic structural realism look significantly limited. Related to this is the even broader point that structural realism essentially applies only to (again, fundamental) physics. There has been little or no effort to unpack the notion of theoretical structural conservatism in other areas of physics, let alone in any of the special sciences, for instance evolutionary biology. The fact that often enough theories in these other sciences don’t look particularly mathematical may be imputed to the relatively early stages of developments of those disciplines, but it may also represent a core limitation of a physics-centric way of looking at science as a whole. Finally, and this is to me a fascinating metaphysical point in itself, ontic structural realism basically collapses the distinction between the mathematical and the physical. Some mathematical physicists, like Max Tegmark (2014) have boldly gone down that route, talking about an essentially mathematical universe. As much as this sort of stuff is fun to think about, it does seem at the least problematic to make sense of the notion that mathematical structures are “real” in an even more fundamental way than physical entities themselves. And so the exploration of the pertinent conceptual landscape continues.
Baggott, J. (2013) Farewell to Reality: How Modern Physics Has Betrayed the Search for Scientific Truth. Pegasus.
Bourget, D. and Chalmers, D.J. (2013) What do philosophers believe? Philosophical Studies 3:1-36.
Cartwright, N. (1983) How the Laws of Physics Lie. Oxford University Press.
Chakravartty, A. (2003) The Structuralist Conception of Objects. Philosophy of Science 70:867–878.
Chakravartty, A. (2004) Structuralism as a Form of Scientific Realism. International Studies in Philosophy of Science 18:151–171.
Chakravartty, A. (2011) Scientific Realism. Stanford Encyclopedia of Philosophy (accessed on 26 June 2012).
Chalmers, D., Manley, D. and Wassermann, R. (eds.) (2009) Metametaphysics: New Essays on the Foundations of Ontology. Oxford University Press.
Churchland, P. (1985) The ontological status of observables: in praise of the superempirical virtues.” In: P. Churchland and C. Hooker (eds.), Images of Science: Essays on Realism and Empiricism, University of Chicago Press, pp. 35–47.
Frigg, R. and Votsis, I. (2011) Everything you always wanted to know about structural realism but were afraid to ask. European Journal for Philosophy of Science 1:227-276.
Kripke, S.A. (1980) Naming and Necessity. Blackwell.
Ladyman, J. (1998) What is structural realism? Studies in History and Philosophy of Science 29:409–424.
Ladyman, J. (2009) Structural Realism. Stanford Encyclopedia of Philosophy (accessed on 16 August 2012).
Ladyman, J. and Ross, D. (2009) Every Thing Must Go: Metaphysics Naturalized. Oxford University Press.
Laudan, L. (1981b) A Confutation of Convergent Realism. Philosophy of Science 48:19–48.
Maxwell, G. (1972) Scientific methodology and the causal theory of perception. In: H. Feigl, H. Sellars and K. Lehrer (eds.), New Readings in Philosophical Analysis. Appleton-Century Crofts, pp. 148–177.
Monton, B. and Mohler, C. (2008) Constructive Empiricism. Stanford Encyclopedia of Philosophy (accessed on 26 June 2012)
Papineau, D. (2010) Realism, Ramsey sentences and the pessimistic meta-induction. Studies in History and Philosophy of Science, Part A 41:375-385.
Post, Heinz R. (1971) Correspondence, Invariance and Heuristics. Studies in History and Philosophy of Science 2:213-255.
Putnam, H. (1975) Mathematics, Matter and Method. Cambridge University Press.
Ross, D., Ladyman, J., and Kincaid, A. (2013) Scientific Metaphysics. Oxford University Press.
Saunders, S. (1993) To what physics corresponds. In: S. French and H. Kamminga (eds.), Correspondence, Invariance and Heuristics: Essays in Honour Of Heinz Post. Kluwer Academic Press.
Smolin, L. (2007) The Trouble With Physics: The Rise of String Theory, The Fall of a Science, and What Comes Next. Mariner Books.
Stadler, F. (2012) The Vienna Circle: Moritz Schlick, Otto Neurath and Rudolf Carnap. In: J.R. Brown (ed.) Philosophy of Science: The Key Thinkers. Continuum, pp. 83-111.
Tegmark, M. (2014) Our Mathematical Universe: My Quest for the Ultimate Nature of Reality. Knopf.
van Fraassen, B.C. (1980) The Scientific Image. Oxford University Press.
van Fraassen, B.C. (1989) Laws and Symmetry. Clarendon.
Whewell, W. (1847) Philosophy of the Inductive Sciences. John W. Parker.
Woitt, P. (2006) Not Even Wrong: The Failure of String Theory and the Search for Unity in Physical Law for Unity in Physical Law. Basic Books.
Worrall, J. (1989) Structural realism: The best of both worlds? Dialectica, 43: 99–124.