Monday, March 16, 2009

Fictions and Realism

The recent interest on fictions and the imagination in science is very encouraging - fictionalists are no longer dismissed as "irrationalists", "relativists" or, worse, "postmodernists". There is however a serious argument that fictionalism is incompatible with scientific realism. It goes roughly as follows. i) The hallmark of a fictional assumption (eg. regarding frictionless planes, point particles, or the elastic solid ether) is the irrelevance of its truth-value to its function in inquiry. So ii) if fictions are indispensible to scientific modeling, then science does not aim at truth - not through its models at any rate. Thus iii) fictionalism - the claim that fictions are rife in science since indispensible in modeling - is inconsistent with scientific realism.

Maybe one should not care about scientific realism, but I suspect this argument is nonetheless fallacious, and I gave a preliminary reply in the introduction to the book I edited for Routledge. There I argued that realists can happily accept a weak form of fictionalism, which endorses what I call fictive representation -- i.e. controllably false idealisation of the properties of real entities. So frictionless planes are ok, point particles may be, but the ether definitely is not. I'm now working on a paper where I claim that even this is too conservative, and that if the right account of fictions is provided then the realist can safely go all the way to the strongest form of fictionalism above. (There is of course a long story behind what I take to be the right account of fictions). The problem lies with the middle conditional premise (ii), but there are few worked out arguments on this topic, which makes it hard to tell what would be standard and what not, so I'd be interested in whatever intuitions people might have (if they have any!).


  1. Interesting. What you seem to refer to can be thought as limit-case thought experiments. Now, can you treat most (some? all? none?) cases of fiction (as used sensibly in science) as thought experiments? If yes, how about taking the Norton-style empiricist view (or Gendler's stand) on thought experiments (and thus on at least some cases of fiction in science?) Do you think this would help with interpreting (at least some cases of) fiction in a realistic manner?

  2. I take it that there is a form of idealization theory that takes these assumptions to be conditionals. For example: if these planes were frictionless, then p, (and p varies continuously with friction, and the planes to which this model applies are very close to frictionless). This assumes a very "controllable" idealization, to use your term.

    So here's the question: are the corresponding conditions for ether theory plausible and interesting?

  3. So here's a thought to get you past the conditional on premise "ii" - Why not restrict the antecedent to the phrase "scientific modeling"? If so, I think that you can start to separate the aim of science from the aim of scientific modeling; it seems as if idealizations are never the case due to the fact that nothing (at least as far as I know) is ever under "ideal" conditions.

    Just a thought.

  4. Hi Mauricio,

    If I wanted to push the anti-realist line, I think I'd point to the link between idealisation and analogies.

    So let's take one of my favourite cases, where a dimer is represented as a set of plates connected by springs (which is in one of my papers on models in physics and biology which I shared with you, I think).

    The anti-realist will point out that this analogy is based on observables (and indeed everyday experience). And if this is generally the case -- there is a leap here, but I can't think of any obvious exceptions -- then this would cast some doubt on full-blooded scientific realism.

    I'm reminded somewhat of Bohr's suggestion that we are limited to our classical concepts, but that these can only go so far, in his discussion of wave-particle duality (or as he would have it, complementarity). Something similar is the case here; instead, we're limited to understanding the unobservable in terms of the observable. Or so an anti-realist might argue...

    Of course, there is no argument here that unobservables are fully comprehensible by analogy with observables. But it's a worry, and there's room for reasonable doubt (if we accept something like van Fraassen's brand of voluntarism), at least.

    (For what it's worth, in my recent work with condensed matter physicists -- and particularly those working with simulations, which always involve idealisations and approximations -- I really did get the impression that an empirically adequate model counted as a successful one!)

  5. P.S. In the first sentence of the penultimate paragraph above, read 'are not fully' for 'are fully'.

  6. Thank you, that's helpful.

    Mohan is right on target: the account of fictions I'm developing takes them to entail conditionals of the kind he describes (NB. I don't say fictions ARE conditionals). The problem is that the idealising conditions in the consequent are not always of the controllable sort - they don't always come equipped with recipes to 'de-idealise'. What's amazing is that whether or not they do seems not to depend on the reality of the putative entity. I had not realised this when I wrote the intro to the book. So, for instance, ether theories stipulate precise procedures to get more accurate rates of em radiation emission and absorption. Point particles are trickier though.

    Rafal rightly points out a connection with thought experiments which often if not always involve fiction. But fictions are much more general and are involved in modeling and real world experimentation too. However, if you have my views on fictions as vehicles for inference then you'd be certainly inclined to Norton's account of thought experiments as arguments.

    Darrell, you might well be right that fictionalism will in the end clash with some kind of realism. I'm not out to defend realism though, but fictionalism. But the strategy is to try to show that fictions are acceptable to as large a range of realists and antirealists (a NOA-like move, as you'll appreciate).

    So the key question for me is related to Mohan's post. How plausible / implausible is the thought that the function of fictions in science is to facilitate expedient inference via entailed conditionals? (NB. no identity claim involved). Nothing speaks against the idea in the literature that I've looked into so far, but then the literature on the topic is nearly inexistent ...

  7. It's pretty plausible, to me at least, that they do fulfil that function (whether or not it's the only function they fulfil).

    Going back to my previous example, one such elementary conditional would be 'If this dimer is squeezed and released, it will vibrate'.

    We'd need to add some caveats, though, perhaps. I think that many fictions are only useful for -- and only employed to help us to get at -- a peculiar class of counterfactuals (relating only to _some_ of the possible behaviour of the entity or entities under consideration).

    For instance, the rigid plates in my example might lead one to conclude that the monomers they map onto do not bend. (So the relevant counterfactual would involve application of particular forces.) But this is false. And this is fine, because the modellers recognise this...

  8. Maybe I should have said a bit more about what I have in mind here. Take some entity. We can use multiple different fictions, in different contexts, depending on which aspects of its behaviour we are interested in. So we can build up a 'piecemeal' understanding.

    I think that this is legitimate science (in so far as we need go no further). I wonder, though, if someone with stronger realist inclinations would agree!

  9. One question and one observation.

    The associated conditional is, as Mauricio rightly says, not the same as the fiction. In fact it can't be the same because

    There are frictionless planes

    is false (hence fictional) and

    If planes in situations S were frictionless, then p, and if certain continuity conditions are met, then close to p when there is friction

    is supposedly true.

    The question is: why does one have to adopt the fiction at all? Or is the claim simply that one is doing certain calculations in a fictional context without necessarily pretending that the fiction is true, and then applying the result to reality via the associated conditional. How would the latter be different from hypothetical reasoning? (I ask because "adopting the fiction" as I call it, is taken pretty seriously in philosophy of literature to explain how we can respond emotionally to fiction. In literature, we are not merely reasoning hypothetically.)

    The observation. In the cases where it isn't clear how to de-idealize (and perhaps Darrell's dimers are examples, though I have not even the slightest notion of what a dimer is), the idealized calculation seems to be a flyer, a shot in the dark. "We don't know why, but this seems to work." Here there is no associated conditional. Here I see better why the fictional approach might be warranted: you just pretend the idealization is true, because you have no idea how to apply it to reality.

    What I am suggesting then is this: in controllable situations, there is no need to fictionalize, but in uncontrollable situations you have no choice.

    Does this make sense, Mauricio?

  10. All we have to do is be possibilists (rather than actualists) and these problems go away, because the alleged "fictions" don't have to be fictions at all! They are real in at least one possible universe.



  11. sciencedefeater: if only it was that easy! But if there's any agreement, fictional worlds need not be possible, the conditionals they license need not be counterfactual, and when counterfactual possible world semantics need not cash them out. (Discuss!)

    Darell: nice example, yes. I agree that scientists often use multiple mutually contradictory descriptions - and indeed a fictionalist explanation is straightforward (more plausible in my view than ontological dappled world' pluralism - would you agree?).

    Mohan: the fiction need not even be false. But certainly the conditionals licensed must be true if the inferences are to be valid - not universally accepted, but a requirement for an inferentialist like me. I don't think though that one can dispense with fictions even when the idealising conditions are controllable. It has to do with the form of the conditionals - not exactly like yours - and their subtle relation to the fictions themselves. I can't go into details here, but just consider the consequences: most fictions in science eventually turn into - typically falsified - hypotheses. Yet, the conditionals remain true. Sure. Vaihinger already knew this. But historically the replacement of fiction by conditionals requires singular vision - only retrospectively does it seem obvious. Think of the ether - it took several generations of physicists and the genius of an Einstein to even come up with the possibility - never mind the detailed list!

  12. Can one be both a fictionalist and a scientific realist? I think that it partly depends on what a fictionalist is. As I understand the label, a fictionalist is someone who think that scientific theories are not to be literally construed. And, I suspect, it is this version of fictionalism that people usually assume (and I think correctly) is incompatible with scientific realism.

    What you call fictionalism in your post, however, does not seem to be incompatible with scientific realism. We may need to use idealized models to apply our best theories to concrete real-world systems because of the sheer complexity of applying them "directly". But, as far as I can see, such idealizations are compatible with the theory being true of the world even if it cannot be applied to the world directly.

    (I suspect that part of the appearance of a conflict is that the scientific realism debate is still mostly discussed in terms of theories and truth, while our most promising picture of science is one in which models and representation play a crucial role. I'm currently working on a paper in which I'm trying to develop a form of scientific realism that is more suitable for this new framework, but I'm afraid I won't have a presentable draft for some time).

  13. Thanks Gabriele, that's an interesting thought. But I differ: if one were to cast fictionalism as an account of scientific theories (and I don't particularly see the point) one would not end up with the view that "scientific theories are not to be literally construed" at all. On the contrary the fictionalist is a semantic realist if there is any. Contrast with either i) classical instrumentalism or ii) verificationism. The former claims theories have no meaning and lack truth values; the latter claims theories' meanings are not literal and their truth-values depend on their relation to protocol sentences. Fictionalism definitely has no truck with either - and if you reflect about these views you will realise that they in fact make fictions literally impossible. (Incidentally 'classical instrumentalism' is also a poor and indefensible construal of genuine 'toolbox' instrumentalism - but you already know my work on the topic ...).

    The paper on realism sounds interesting - keep it up!

  14. Hi Mauricio,

    Let me try to state this more carefully. As I understand it, fictionalism as a position in the sci realism debate, claims that the aim of scientific theories, just like the aim of works of fiction, is not that of describing the actual world truthfully (even if it accidentally theories and ficitons might happen to describe the world truthfully).

  15. Gabriele,

    I think we have to be careful to distinguish between theories and models. The models can be fictions (or based on fictions) when the theories/laws (e.g. qua universal statements) are true, for instance.

    I'm also not sure what exacty 'the aim of scientific theories' is supposed to mean. Is this distinct from 'the aim of science' in the standard sense, e.g. of Popper and van Fraassen? (Are you suggesting that theory-construction, say, has an aim which may be distinct from the aim of science generally? I don't agree 'the aim of theories' can simply be mapped on to 'the aim of science', if that's what you're suggesting. Science involves so much more... and may, of course, have multiple non-hierarchical aims.)


    Yes, I think a fictionalist explanation is more plausible than the alternatives I'm aware of.

  16. Darrell,

    That's the distinction I was trying to make in my first comment.

    And, no, I meant it in the same sense as van Fraassen but I'd rather prefer scientific realism and its rivals to be formulated in terms of epistemic justification rather than the aims of science.

    Hope this clarifies.

  17. Hi Gabriele,

    At the risk of going off-topic, can't a link between method and aim be required, with respect to 'what counts as success in science', without requiring that we discuss matters in terms of epistemic justification?

    (It's generally not reasonable to suggest that X is the aim of some activity if X can't be achieved by performing said activity. And so on... I discuss some of this is my forthcoming AJP piece, where I cover evolutionary arguments for the aim of science. Amusingly enough, I conclude that just about the only defensible aim, on such a strategy, is to rule out empirically inadequate theories. So I think that both Popper and van Fraassen argue for something too strong! Especially as they are both, like me, anti-inductivists.)

  18. yes, with Darrell's caveats that woudl be the standard view nowadays, I think. But I was not arguing at all against Gabriele's understanding of fictionaling as idealising in the application of models (which is roughly how I too like to think about it) - and I very much look forward to his paper. Rather I took it that Gabriele was making an interesting point about the source of some philosophers' irrational fumings against fictionalism (but this seems to be a thing of the past since interest on the topic is very visibly up these days) - namely that they may have mistaken it with a discredited form of semantic antirealism. And we agree that if one chose to understand fictionalism as an account of scientific theories (which neither of us three seems tempted to do) one certainly would not end up with anything like semantic antirealism. Thus the objections are besides the point, and the source of the resistance must lie elsewhere. Where? In some naive and unreflective 'scientism' perhaps?

  19. Another possibility may be that different notions of 'theory' are operating...

    Let's imagine I just said "I'm a fictionalist about theories" to someone who took putative law statements like "All bunnies are brown" to be theories. It's not hard to see that they might take me to be a semantic anti-realist.

    (How would one explain the position to such a person? I'm minded of the way that regularity theorists like Ayer explain Newton's first law as a special case of the second in response to the suggestion that the former is vacuously true. Here's one example of a fiction, namely a body without any forces acting on it, having been put to good heuristic use. And we definitely want the 1st law to be taken just as literally as the 2nd!)