Harman doesn’t understand structural realism

Posted by | peter | 28.3.13 | 2 Comments

A review of Harman (2010) “I am also of the opinion that materialism must be destroyed”, in Environment and Planning D: Society and Space, vol. 28, pg772-790  

I have finally got round to reading Harman’s engagement with structural realism. It is an attempted exegesis of Ladyman & Ross (2007) Every Thing Must GoMetaphysics Naturalised (henceforth L&R and ETMG reciprocally). The article is replete with misunderstandings and although I normally refrain from commenting on OOP, on this issue I really can’t keep quiet. All references are to this paper unless stated otherwise. And the following acronyms are used:

OSR – ontic structural realism
MSR – moderate structural realism
ESR – epistemic structural realism
QM – quantum mechanics
QFT – quantum field theory
OOP – object-orientated philosophy
PII – The Principle of the Identity of Indiscernibles

(sorry there are so many but I’m lazy). Here are some (mostly negative) comments, in no particular order:

On misunderstanding causation
L&R’s engagement with the concept of causation is complex but Harman manages to gloss over its intricacies with ill-concealed prejudiced language.  L&R analyse causation as a threefold concept: a) the folk concept of the manifest image – what Scott Atran calls part of folk-mechanics that we use to track physical reality at the speeds in our manifest image of the world; b) the notional concept of causation which is used in the special sciences to track representational real-patterns; and finally c) the denial that causation is a fundamental property of reality because fundamental physics is better conceived of in terms of “interactions”. Reciprocally, Harman calls these a “folk product” (778), a “concession” of the special sciences (780); and non-existent at the fundamental level (but his tone here smacks of disbelief).

Although Harman has this correct, the way he describes it belies a lack of understanding on his part. Rather than a “concession”; L&R’s notional concept of causation follows directly from their naturalistic principles and their acknowledgement that non-reductionism is the appropriate position for such a metaphysical project. Harman’s misconstruction becomes apparent when he states that L&R’s “ambiguous” position on causation means that “it ought to make it difficult for [a] table to […] block our progress […] if we stumble into it” (782). L&R never deny specific interactions, they just deny an overall metaphysical causal concept for all cases – the sought that Harman would like. 

Furthermore, this error is compounded when Harman goes on to imply that if causation is not a fundamental structure of the world this makes L&R equivalent to occassionalists (779). And subsequently states that real patterns are “discrete” (782). Since L&R deny outright the existence of independently existing and subsisting objects – as the sections on fundamental physics and the metaphysical underdetermination of the non-individuality of quantum particles, spacetime points and particles and fields in QFT makes clear – this comparison is a complete non-starter (see ETMG chapter 3). The only way Harman could have drawn such a comparison is if he didn’t understand the empirical evidence put forward from fundamental physics. This is quite damning, since this is a key component and motivating factor of OSR – as stated on multiple occasions by Ladyman, French and other proponents of OSR (see, amongst others, ETMG, chapter 3; French & Ladyman (2003) “Remodelling Structrual Realism: Quantum Physics and the metaphysics of structure”, (2011) “In Defence of Structural Realism”; and French (2010) "The Interdependence of structure, objects and dependence").

Harman’s brief discussion about PII and Pound Sterling confirms this suspicion (782). Harman states that PII holds in something he calls “classical metaphysics”, which he does not define, but could be interpreted to mean metaphysics devoid of contact with what the physical sciences tell us about reality. He goes on to state that PII does not hold in mathematics and quantum theory because “the relational properties of both Pounds are the same”. Firstly, whether or not PII holds in mathematics is irrelevant since we are in interested in physical reality and not logical reality. Secondly, ignoring the fact that quantum mechanics does not straightforwardly apply to pound coins since they don’t strictly exist at the microscopic scale (this manoeuvre is how L&R and also separately David Wallace attempt to diffuse the measurement problem – for more information see EMGT, 180-183, esp.182); it is not that the relations of quantum particles are all equivalent that renders PII physically fallacious. It is that quantum particles are not clearly individuated in all events such as we would expect if PII held and if straightforward logic held. Instead, fermions and bosons both behave in specific ways that make the concept of “individual” suspect – but rather than then stating that quantum particles are not individuals. OSR is premised on the metaphysical underdetermination of multiple ways of characterising particles that are co-existent and not easily chosen between: haecceity; non-individuals; relationals; contextual individuals. This situation leads proponents to argue that rather than an end of metaphysics – as constructive empiricists suggest – it is suggestive that the metaphysical concept of individual is anthropocentric and inappropriate to mind-independent reality (see French & Ladyman 2011 and ETMG, 132-140). Considering that Harman is after a non-anthropocentric metaphysics, it is a shame that he misses this central point of OSR and the book.

In fact, Harman seems incapable of understanding this point, and the way he describes L&R’s rejection of individuality implies that he seems to think it is an a priori metaphysical prejudice (783) rather than dependent upon empirical evidence. And this is excellently apparent when he writes: “The latter point [that QM suggests there are no objects] is easily disposed of by noting that this is by no means the universal interpretation of quantum theory” (785). This is exactly the point of their overview of the various metaphysical interpretations of individuality/non-individuality problems and the metaphysical underdetermination this creates. All I can say is that Harman needs to reread chapter 3, the empirical evidence is important.  

Furthermore, other structural realists argue that the real patterns that fundamental physics tracks must be thought of as causal structures (see for example, Esfeld and Lam (2009) “Structures as the objects of fundamental physics” and Saunders (2003) “Structural realism, again”). This issue depends upon whether the universe is dynamical or non-temporal – see next comment.

On speculative questions
Harman states that L&R should have a problem with plurality and require pre-existing individuals (782-783). On this point Harman should re-read section 3.5 (ETMG, 154-159) where he would see that this is a debate within structural realism about the need for relata between relations. French & Ladyman 2011 explores this more, and shows that OSR accepts nodes or thin notions of objects – objects entirely defined by their relations.

Harman’s problems here about how these structures “surge into existence from an incompletely differentiated structure” are interesting. And I would suggest that people interested in Iain Grant’s work would also find the puzzle of the generation of structure fascinating. On this note I suggest Lee Smolin and Julian Barbour’s work in seeking a mathematical formulation of Leibniz’s monadology – in which they are trying to articulate “a structure-creating principle” (see Barbour (2003) "The deep and suggestive principles of Leibnizian philosophy" for a less technical examination of this). But L&R, for better or for worse, explicitly reject such metaphysical questions as overly speculative – and so asking them this question is bound to be met in silence. They couch all of their metaphysical speculations in terms of the problem of quantum gravity (see ETMG, 167-175) since this will go a long way to deciding whether we live in a block universe or one in which dynamical collapse and thermodynamics are fundamental. L&R side with the block universe, but this is not central to the thesis of the book. So a suggestive answer to Harman’s question about “surging into existence” is that this is a meaningless question since time and causation are not fundamental, and instead are indicative of philosophising from within the manifest image. A mistake Harman repetitively makes – whilst proclaiming the opposite. And this is my main problem with OOP. Rather than being non-anthropic, its one of the most extreme forms of anthropocentricism you can get – viz. just because objects recede from our attention in the manifest image does not mean that other objects recede from each other when they interact. Such a statement is a straightforwardly anthropic projection: “the world is like this for me, therefore it is like that for things different from me”. Literally, Spinoza’s “god is a triangle”.     

Beyond this, I struggle to actually understand why Harman thinks that L&R cannot have multiplicity (787-788). And his discussion of whether reality is discrete or continuous is a centrally vexing issue in the pursuit for quantum gravity – as L&R acknowledge, and hence, given their naturalistic principles they remain mute on a final deciding vote in this dilemma whilst also exploring the issue and laying out all the various angles to their full extent (see the section on Quantum Gravity in ETMG). I struggle to understand how Harman could want more at this juncture given that physicists have been struggling with this problem for nearly a century.   

On empirical falsification
In answer to Harman’s query about how the real-patterns conception of reality could be empirically falsified (780); if it could be shown that one of Harman’s bizarre conglomerate substances – such as “my left nostril and the capital of Namibia and Miles Davis’s last trumpet solo” (ETMG, 231) – exists, something that L&R deny, then this would demonstrate that their existence claim “to be, is to be a real pattern” is false. What is funny about this collection, as are all of Harman and Latour’s collection substances, is that they are entirely human centred. Entirely. In contrast, L&R’s conception of real-patterns, especially extra-representational real-patterns, fit perfectly into the non-anthropocentric science that is actually being carried out by machines. Humphreys (2004) excellent book, Extending Ourselves, outlines how the exponentially increasing computational paradigm has rendered anthropocentric stances, both anti-realist and realist, null-and-void. Humphrey’s work is also important considering our next issue.

On accusations of correlationism and repetitive pronouncements of the inexhaustibility of objects
When Harman states that there can be nothing outside of human or animal observation for L&R (784), he clearly didn’t read the bits discussing Humphrey’s work on how most of the experiments of fundamental physics are conducted by machines. That would be machines and particles or quasers interacting – devoid of human or animal mediation. So the statement that: “Everything boils down to a correlation between physical structure-in-itself and mathematical structure-for-living-creatures” is simply not true (784). And the amount of errors in Harman’s exegesis is making me wonder how this article was peer-reviewed.   

Along these same lines, when Harman discusses Neptune (786-787), it must be noted that just because Neptune is more than its relations with our best current science does not mean that Neptune is more than its relations with the rest of reality. Nor does it mean that this leads to a difference between relations and relata. It does not matter how often Harman states that objects are more than their relations, the evidence is against this view. And whilst there is a reasonable dispute between MSR and OSR about whether there are relata as relationals or mere nodes, there are not distinct haecceities since this contravenes the diffeomorphism invariance of general relativity and permutation invariance in QM (see French & Ladyman 2011 and ETMG, chapter 3). Although it should be noted intrinsic identity cannot be strictly ruled out by empirical evidence, since it is the claim that it is a transcendent individuality intangibly over and above the empirical. The fact that it contravenes these symmetry translations just heavily implies that it is false and that a more parsimonious view is that haecceities do not exist. To repeat: OSR’s prodecure is to remain agnostic in this debate since it is anthropocentric and point out that fundamental physics operates perfectly well without this folk-metaphysical category. Instead, it can be couched out well in terms of structures – although whether these are dynamical or causal is an interesting and ongoing issue (see first comment above).

In fact, Harman’s repetitive pronouncements for the necessity of objects are tiresome and his statement about the “futility of all other options” couldn’t be further from the truth (788). Chapter 3 of ETMG lists a panoply of reasons for adopting a relational metaphysics and the exploration of reality carried out in fundamental physics works just fine without any thick notions of an object that Harman would like. That Harman is unaware of this after having apparently read the book is amazing. It’s not even as if he disagrees; his statements aren’t disagreements and rebuttals, they’re oblivious proclamations.        
One last note on this subject: on how to actually overcome the correlationist gambit, see Brassier’s (2010) “Concept and Object”, and in particular his discussion of Saturn. It’s really good and shows how to cut through the tautologous gem-like “to think X is to think X…” etc. with a precision and eloquence that is delightful (and, in distinction to OOP, actually works).

On the relationship of the mathematical and the physical
Harman is right to pick out that the peculiar relationship of the mathematical and the physical is not clarified (785). However, it is far from clear that this distinction is the “central distinction of [L&R’s] philosophy”. I would argue that the central distinction of their philosophy is between naturalistic, scientifically motivated metaphysics and scientifically uninformed, neo-scholastical metaphysics.

I wrote my masters dissertation on just this issue and I hope to pursue a PhD in carrying on my investigation – what Mark Colyvan calls the “Wigner Puzzle” (2001, “The miracle of applied mathematics”). I can see why L&R doesn’t want to tackle such a question, since they will see it as going beyond the verifiable remit they laid down as one of their core principles. However, I do agree with Harman that this question is not properly engaged with. But it is not resolving this issue that is at stake, but rather its exploration that is necessary, otherwise OSR is open to the same charges that Ladyman (1998, “What is Structural Realism”) lays at the feet of ESR – namely that it is an “ersatz realism”. Furthermore, I think this question or puzzle can be engaged with in a manner consistent with the principles of naturalistic metaphysics. And given Mark Steiner’s (1998, “The Applicability of Mathematics as a Philosophical Problem) assertion that the Wigner Puzzle implies anthropocentricism, a “user-friendly” universe and a challenge to naturalism, this question becomes even more pressing. 

On this note, when Harman discusses the isomorphism of the physical and the mathematical in the conclusion; although an isomorphism is a “translation” (788-789) I’m not sure about Harman’s Latourian terminology here (I’m sticking with the normal mathematical usage). Isomorphism is a one-to-one correspondence between two structures A and B. Correspondence is not the thing itself, it is a type of relation that suggests an identity of structure in the two separate relata. Therefore, Harman’s comments about knowledge as the thing-itself here miss their target. L&R are not attempting a mathesis universalis of the type attempted by Leibniz and Descartes where the idea was the thing. Instead, OSR is premised on the isomorphic or partially isomorphic relationship between our mathematical models and the world structure itself. This is a normative relationship about a modal structure – and this is the extent of their realism: that there is a real, mind-independent, modal structure, that we call nature. Our relationship to this is best explored through mathematical models – this is a fact. How it works exactly is the Wigner Puzzle, but it is not to mistake knowledge and the thing that the knowledge is about. Again see Brassier (2010) for more on this topic.  

Apart from this misconstrual, Harman is correct to say that there is an issue here in the relationship between the mathematical and the physical. However, this issue is far from the death-blow he believes it to be. And is simply an interesting puzzle that requires further exploration.      

On Naturalism
Harman states “It is not clear why philosophers must prematurely unify their own speculations on space, time, and substance with those of quantum theory and relativity that are not yet even unified with each other” (785). Firstly, does Harman know why these fields in physics are not unified? I think this is an important question, since for him to hold such an aloof attitude as to what metaphysics should be allowed to do and ought to do, becomes interesting in terms of how scientifically informed he is. Viz. to hold a dismissive attitude about something is fine if you actually know what that thing is. If you don’t know what that thing or enterprise is, then your dismissal is potentially wrong, and you won’t even know it. Secondly, purely a priori speculations about space and time devoid of relativity are anthropocentric. Physics tells us how space and time behave and there are a whole host of really interesting metaphysical problems here such as the hole argument (see the entry in the Stanford Encyclopaedia). OOP can’t discuss this physically meaningful metaphysical problem because it is too busy playing around in the logical phase-space of the metaphysical playground. It’s imaginative and fun, but it isn’t about how reality actually is. Thirdly, findings from fundamental physics suggest that substance is an obsolete metaphysical category, and only relevant for couching out our folk mechanics in the manifest image. As Poincare wrote: it is a crutch for our weak anthropocentric minds.       

If you want to do metaphysics properly then you must pay attention to what the requisite physical sciences tell us about nature (see ETMG, chapter 1).

On science influenced by philosophy
Lee Smolin is probably the most famous, to my mind, scientist calling for philosophers to influence science in these troubling times of trying to tackle the intertwined problem of quantum gravity. Harman relishes in this sort of territory. He lists Einstein, Mach and Bohr as influenced by philosophers, such as Kant, and then cites Leibniz as being involved in the physics of his day (785). This statement is fine until you realise that a) both Kant and Leibniz where scientifically literate in the physics and mathematics of their day; b) properly philosophical speculation at the boundaries of our knowledge of reality is being done by people informed by our best theories about that reality (namely fundamental physics – see for example Roger Penrose’s most recent popular work, or David Wallace (2012) The Emergent Universe).

Citing that Carlos Rovelli wants more philosophers involved in the problems of quantum gravity does not override these facts (786). Especially given that in Quantum Gravity Rovelli states quite clearly that nothing has ever been achieved by uninformed speculation: “Wild ideas pulled out of the blue sky have never made science advance” (2003, 5). So when Harman comments that Ladyman gives a “weak answer” (786) to Rovelli’s call to arms because he says that currently only a few philosophers are informed enough to work at the “cutting edge” – such as David Wallace, it’s a little bit misleading. To be fair to Harman, his reasons for thinking this is a weak answer are that Rovelli isn’t calling for philosophers to be on the frontlines of science, but to be “beyond the cutting-edge” (Harman’s emphasis). This would be a somewhat good point. But given Rovelli’s actual views we can ask: if philosophers do not even know where the cutting-edge is how the hell are they supposed to roam around in this speculative territory?   

a) I am struggling to see why Harman thinks L&R’s stating of the difference between Kantianism and OSR is just “table pounding” (783-784). The former says science cannot access noumena; whereas the latter enquires into the real structure of the world. I think that this is quite a clear difference and also draws the distinct between ESR and OSR – a distinction, made clear in chapter 2 of ETMG. And several papers have been written about how ESR is Kantian and OSR isn’t – especially given French and Ladyman’s admiration of the neo-Kantian Cassirer (see 2003). 

b) Harman’s statement that L&R’s approach, of combining verificationism and realism, is nothing new is rendered obsolete considering that he doesn’t even understand what their position is (784). And that is putting it politely.

c) Against Harman I would say that stating that intuition is a socio-biological phenomena is not a “red-herring” (786). Identifying and removing intuitions from metaphysics is how we avoid anthropocentricism. The critique of metaphysics helps us in this task as Peter Wolfendale has brilliantly demonstrated in his essay on Transcendental Realism (and as Schelling rather differently puts it in the introduction to his Ideas for a Philosophy of Nature).    

d) It may seem trite but there is at least one point on which I sort of agree with Harman: “A truly multipolar cosmos requires that the human being be treated as just one kind of entity among trillions of others, not as a full half of a dual monarchy” (772). I completely endorse this sentiment, if not its peculiar phrasing; but only with the caveat that the human brain is the most complex entity we have encountered in the universe. This goes some way towards explaining the abundance of creative relations and activities with which it interacts in the universe and why there are more sciences involved in investigating these than say the various branches of geology – viz. rocks are not as complex as brains. This is not to say that one is more special than the other, just to point out that one has more affective relations in nature (as identified as early as on as Spinoza I would argue). To view them as both “actants” is a far worse form of reductionism than any scientifically motivated metaphysics as Brassier (2010) demonstrates clearly.

On Materialism
Given the title of the paper, this is a central point. As such I have left it till last and as the conclusion of these comments. Harman defines materialism as “a standpoint that breaks down the dualism of subject and object, allowing these two poles to interpenetrate and mutually constitute one another” (772). I am not sure that L&R would recognise this definition since they are seeking a middle ground between van Fraasen’s empiricist and materialist stances (see Ladyman 2011 "The Scientistic Stance" and ETMG, the end of chapter 1), which are more to do with attitudes than substances of subjects and objects. It is likely that they would see this latter issue as a neo-scholastical one: the world is not made of anything.

So when Harman states that “materialism must collapse into object-orientated philosophy” (789), I am sure he is correct for his own definition. But since it is not clear whether it has any bearing on what L&R are talking about, it is meaningless assertion. If you choose to define your opponent in an idiosyncratic way and then undermine this definition, your opponent can merely say “that’s not me” and ignore what you’ve said. 

A positive general point that Harman draws out is that materialism and idealism are quite similar positions in certain ways. A point that he has made before, and so have others. And so his assertion that materialism may not be the best name for a scientifically informed naturalism does hold some water (it’s not an entirely leaky boat). But I think this point is not as important as naturalism – which he does not touch upon sufficiently.

And all this obviously reflects badly on the title – although I did enjoy how humorously violent it is and it reflects Harman’s excellent skills as a communicator and creator. At this juncture I would like to state that Graham Harman’s writings (along with those of Iain Grant, Ray Brassier and Quentin Meillassoux) is partially what recaptured my desire to do philosophy. And I admire his ability to create some excellent philosophical tools and his superb turn of phrase (“rainforest scientism” (779) is particularly great). Beyond this, however, I do not think OOP is a properly credible philosophical position – for more details Peter Wolfendale’s essay on the failings of OOP. And I have nothing further to say on the matter except that anyone who actually wants to know what structural realism – probably the most sophisticated form of scientific realism – is, would be better off reading the entry in the Stanford Encyclopaedia or Frigg & Votsis (2011) “Everything you always wanted to know about structural realism but were afraid to ask” or ETMG itself.


2 Responses to “Harman doesn’t understand structural realism”

  1. UWE Grad
    29/3/13 04:31

    I really enjoyed reading this Alex, not least because it's an interesting document of the direction of your research interests at the moment. I remember you reading ETMG over the summer! Who knew where it'd take you?

    One question I'm particularly interested in is the relation or priority of empirical facts and metaphysical speculation. I'd imagine the structural realists have quiet a lot to say about this. Are there cases when metaphysical necessity might trump empirical contingency? Would this be the case in paradigm changes? Or is it ever possible to argue, for example, that metaphysics of objects is necessary and something must be done to bring science in line with this demand?

  2. peter
    13/4/13 18:45

    I think the structural realists are quite divided and varied on the relationship of metaphysics and science, although they roughly all oscillate around some form of naturalism.

    I'd say that the relation of metaphysics and empirical facts is reciprocal. I re-recommend Peter wolfendale's essay on transcendental realism which covers this topic really well. A somewhat rehashed retelling of some of his points is: the relationship is something like Quine's web of beliefs - all the facts and suppositions are interconnected in a certain manner and all are up for grabs and examination. Certain metaphysical propositions will sit more firmly embedded within the framework and ground a lot of other facts and details, but even these are also up for revision and examination.

    JH Spencer has argued that there are certain Platonic metaphysical necessitites for science: unity, symmetry, order, beauty. But I think he has drawn an arbitrary line - even these can be examined and questioned and found wanting.

    In contrast to Spencer's attempts to ground science in a fideistic faith (like Kant's principle of purposiveness in the CPJ), I think that each of these principles can be questioned. The mistake that Wolfendale identifies within these fidestic approaches is that they mistake the inability to question all of one's principles simultanously ("globally" - as he puts it) with the inability to question the deep structures of our metaphysical or transcendental framework at all. Viz. we can examine this structure locally but not globally - e.g. we can look at the concept of unity and its empirical verifiability, but not whilst simulteanously questioning a whole host of other transcendental conditions that legislate whether we can think at all.

    With regards to objects, I think the empirical evidence is in: http://www.unil.ch/webdav/site/philo/shared/DocsPerso/EsfeldMichael/2009/Structures-MPI-09.pdf

    this paper isn't very long and it summarises the empirical evidence (from QFT, GR and QM), gives nice overviews (and references for further details) of the arguments about structural realism and also explicitly engages with the metaphysical outcomes of the scientific evidence. Its also really interesting because they (Esfeld and Lam) argue that numerical diversity pops out of the empirical data as a metaphysical necessity. Would love to know your thoughts on this Chris.

    I suppose if a certain form of hidden variables turned out to be true rather than dynamical collapse or many-worlds, then this would imply a set of values beyond the presently relational set that we currently have. But then the argument would again be whether these are haecceities (or non-relational properties that OOP craves) or again that these are just further purely relational aspects of purely relational thin-objects. My hunch is that even if a version of hidden variables is true, that the latter is more likely given the empirically well verified findings of QM, QFT and GR and the violations that haecceities cause with these empirical facts.

    sorry for such a long answer!

Leave a Reply