Hijacking Correlationism

Graham recently put up an interesting post about the various positions within Meillassoux’s philosophical ‘spectrum’, and where OOP stands in relation to them (here, linked by Gratton here). This is most interesting, because it goes some way to confirming the diagnosis of OOP I made in my TR essay (here). Since most people don’t have the time to read the whole thing, I’ll recreate the basic elements of the argument here (with a certain amount of tweaking).

First of all, the core point of the essay is that the ‘spectrum’ of positions provided by Meillassoux is incomplete, and that there are at least two further important positions (not including OOP) that need to be added to it, which I call deflationary realism and transcendental realism. The revised range of possible positions should be something like: classical realism (Aristotle, Locke, etc.), classical idealism (Berkeley, Hegel, etc.), weak correlationism (Kant), strong correlationism (Wittgenstein, Heidegger, etc.), speculative materialism (Meillassoux), OOP (Graham, and perhaps related OOO variants), deflationary realism (Quine, McDowell, Brandom, etc.), and transcendental realism (me and potentially a few others). I won’t line these up in a spectrum, because I think there’s too many dimensions at work here.

The other relevant point that I made is that both Meillassoux and Graham justify their respective positions by hijacking the arguments for correlationism, albeit in different ways. This is very explicit in Meillassoux’s work, though has been somewhat more understated in Graham’s (although his post makes this explicit to some extent). On the basis of this, my argument was that if we undermine the arguments for correlationism directly, then we undermine the most powerful arguments in favour of both speculative materialism and OOP. This was then done by showing that despite the fact that correlationism is meant to be an epistemological position (or at least that we are supposed to be able to formulate it in purely epistemological terms), it depends upon certain implicit ontological (and thus metaphysical) assumptions. In effect, what Meillassoux and Graham do in hijacking correlationism is just to try and make these assumptions explicit, and work out their consequences. The problem is simply that once one recognises this, one sees that these are not metaphysical positions that are necessitated by non-metaphysical (epistemological or phenomenological) facts, but are just different ways to develop some existing metaphysical assumptions. Arguing against those assumptions thus undermines correlationism, speculative materialism, and OOP all at once.

This is a very schematic presentation of these ideas, which doesn’t show how the two sides link up. As such, I’m going to try and flesh it out a bit.

1. Realism, Idealism, and Correlationism

The first thing to do is to understand what is going on in the debate between classical realism and classical idealism, and how correlationism fits into this dialectic. In the essay, I’ve put forward some fairly condensed definitions of all three positions. I’ll start with the first two:-

Classical Realism: Any position that takes there to be a real structure of the world that is ontologically independent (and thus distinct from) the structure of thought.

Classical Idealism: Any position that takes there to be a real structure of the world that is in some sense identical to (and thus ontologically dependent upon) the structure of thought.

The difference between these positions has often been understood as a matter of identity and distinction, although the terms which are supposed identical or distinct differ depending on the formulation. The choice between these positions is often understood in terms of whether there is an identity between thought and Being, subject and object, knowing and known, and the for-us and the in-itself. Although they are related, these descriptions aren’t exactly equivalent, insofar as on the side of thought (or knowing) we can draw distinctions between individual acts of thought, the contents of these acts, the subjects which perform them, and the structure of thought as such, and on the side of Being we can draw distinctions between aspects of particular objects (or entities), particular objects, the structure of objects as such, and the structure of objects as a whole (or the structure of the world). It is wise not to collapse these distinctions, as doing so leads to confusing various positions that are distinct. For instance, if we fail to distinguish between the subjects, acts and contents of particular thoughts and the structure of thought in general, then we will fail to distinguish idealisms that hold that the thought of an object is in some sense identical with the object (e.g., Berkeley’s empirical idealism – esse est percipi) from those which simply hold that the structure of thought is identical with the structure of the world (e.g., Hegel’s absolute idealism – the identity of Being and Concept). This is why I have defined classical realism and classical idealism in terms of the relation between the ‘structure of the world’ and the ‘structure of thought’, because it represents the real fault line between the various idealist and realist positions. Berkeley and Hegel might disagree on the identity of particular thoughts about objects and the objects they are about, but they nonetheless agree on the identity of the structure of thought and the structure of the world (at least to some extent).

The really interesting issue here is whether we take ‘Being’ to mean ‘the structure of entities (or objects) as such’, ‘the structure of the world’, or both. This is a crucial question which is of core importance for understanding Heidegger’s work, and the historical relation between ontology and metaphysics. The answer is complicated, but I won’t provide it here, as I’ve written about it more in the TR essay and elsewhere (here).

The other issue with these definitions is one raised by Jon Cogburn, who pointed out to me that there is a legitimate question as to whether Spinoza and Leibniz are classical realists or classical idealists on these definitions. This is because there’s a genuine question over whether the structure of thought is adequately shared between us and God. Of course there is some genus ‘thought’ to which both man’s thought and God’s thought belongs, which plays a crucial role in propping up the principle of sufficient reason in each case (which I’ve written about here), but there are also some concrete differences in each case, necessary to make the whole thing work (i.e., the infinity of God’s thought as opposed to man’s finitude). In essence, I think they both hold that there is some ontological dependence between the structure of thought and the structure of the world, but that this is not straightforwardly a matter of identity. Whether or not one wants to read them as more realist or more idealist ultimately depends upon whether one emphasises the status of God as a feature of the structure of the world, or as a thinker in some sense comparable with finite thinkers.

Moving on, we can define correlationism as follows:-

Correlationism: Any position which holds that the real structure of the world is to some extent unknowable, insofar as knowledge is always relativised to the subjective conditions of knowledge (e.g., forms of intuition, cultures, or language-games, etc.). This means that there can be no access to the world as it is in-itself, but only as it is for-us.

From these definitions it is fairly obvious that classical realism and classical idealism are defined in ontological (and thus metaphysical) terms, but correlationism appears to be a purely epistemological position. In order to see how it is implicitly ontological, and to understand how it is hijacked by Graham and Meillassoux, we need to delve into the arguments justifying correlationism.

2. The Dialectic of Correlationism

Meillassoux does a very interesting job of reconstructing the dialectic between realism, idealism and correlationism, in order that he can hijack it. The basic idea is that the correlationist has two arguments: the circle of correlation against realism and the argument from facticity against idealism.

The circle of correlation maintains that one cannot think X without thinking X, or that one cannot think anything as it is independent of thought. Meillassoux finds the simplest form of this argument in Fichte, in the stripped down conception of consciousness that he provides at the beginning of the Wissenschaftslehre. This argument is called a ‘circle’ insofar as any counter-argument the realist provides can always be relativised to the structure of thought, as if we added ‘for-us’ to the end of all their sentences, making the argument go round in a circle indefinitely (A: P, B: P ‘for-us’; A: No, because Q, B: Ah, but Q ‘for-us’; etc.).

Now, the idealist can accept the result of the circle of correlation, but simply maintain that the ‘in-itself’ and the ‘for-us’ are identical (although this may be cashed out in different ways). In Meillassoux’s terms, the idealist absolutises the correlation. To refute the idealist, the correlationist needs to shore up the possibility that the in-itself is different from the for-us. This is done by insisting on the facticity of thought. This is to say, the existence of thought is taken to be contingent. Given that there must be a real structure of the world, this implies that there can be a world without thought, and thus that they are not identical.

The important thing to recognise is that this dialectic is structured around two different questions about dependence upon the structure of thought:-

1) The Ontological Question: Is the in-itself ontologically dependent upon the structure of thought?

This is the primary dimension of the debate between classical realism and classical idealism as I have sketched it. Correlationism sides with classical realism on this issue, holding that the in-itself cannot be dependent upon thought because of its facticity.

2) The Epistemological Question: Is our knowledge of the in-itself epistemically dependent upon the structure of thought?

This is the debate over whether we can know anything absolutely, or whether all knowledge is relative. Both classical realism and classical idealism are forms of absolutism, whereas correlationism sides with relativism on this issue. The problem here is that idealism can accept that knowledge of the in-itself is in some sense dependent upon the structure of thought, because it accepts that the in-itself is ontologically dependent upon it, but nonetheless deny any form of relativism. In short, this question must not only be whether there is epistemic dependence, but also whether it is radical epistemic dependence. To establish the kind of radical dependence it wants, it is necessary for correlationism to provide a negative answer to the first question (so, realism = no, no; idealism = yes, yes, and correlationism = no, yes).

However, the crucial point is that the correlationist also needs to show that the ontological independence of the in-itself from thought is sufficient to establish the possibility that the in-itself could be different from the way it appears in relation to thought, and thus to establish radical epistemic dependence. In the essay I called this the sufficiency thesis.

Now, I should qualify this thesis by saying that the fact of ontological independence is not necessarily meant to establish radical epistemic dependence all on its own. Rather, the idea is that the circle of correlation establishes some form of epistemic dependence that the fact of ontological dependence is sufficient to upgrade to radical epistemic dependence. In essence, Meillassoux’s correlationist is claiming is that the very structure of thought implies that there is some kind of dependence of our knowledge of the in-itself upon its subjective conditions (circle of correlation), and that because the in-itself is ontologically independent of thought (argument from facticity), this means that the in-itself could always be otherwise than our knowledge takes it to be, thereby making our knowledge not just dependent upon, but relative to those conditions. However, this inference is not obviously a good one – the sufficiency thesis is open to question.

The reason for this is that precisely what kind of epistemic dependence the circle of correlation is supposed to establish, and what this dependence consists in, is never specified. A response to this might be to point out that this is because Meillassoux has only provided us with a form of argument, rather than a complete argument, and that different correlationists fill it out in different ways by providing their own specific accounts of epistemic dependence. This is all well and good, but what it indicates is that the simplistic form of the argument must be supplemented. Ray Brassier has gone to some lengths to show the weakness of the circle of correlation by showing how it is equivalent to Stove’s gem, which is in effect an attempt to derive something from a tautology (which is always fallacious). I won’t go into too much detail here other than to say that I entirely endorse his analysis. The real question is whether fleshed out versions of the correlationist argument, which don’t proceed by the threadbare argumentation of the circle can hold up, and if so, how they hold up. In his paper from the TR workshop, Tom O’Shea has done a some pretty good work along these line, by showing why, even if he is a correlationist, Kant’s arguments for his position can’t be reduced to the circle of correlation.

Leaving Kant aside, the question is this: how must the circle of correlation be supplemented in order for the sufficiency thesis to hold up?

The first point to make is that the simple form of the circle depends upon a very stripped down account of thought, and that part of the reason this account is so thin is so that it can potentially encompass a variety of different more determinate positions, such as Kant and Fichte’s views of judgment/positing on the one hand, and Husserl, Heidegger and Derrida’s views of experience and language on the other (not to mention the later Wittgenstein’s linguistic account of thought). In the TR essay, I’ve gone into a bit more detail about the particular distinctions that the account overlooks, but again, I won’t rehearse them here. However, it is important to note that it doesn’t pay any attention to the distinctions between subject, act, content and structure that I made above. The reason for this is that without making these distinctions it is hard to understand precisely what the terms of the dependence relation are.

The implication is that it is the content of our knowledge which is somehow relative to its subjective conditions (which at minimum involve the structure of thought in general, but could include the particular constitution of the subject and the particular circumstances of the act). The basic idea here is one that should be reasonably familiar, namely, that the content of our knowledge (or the for-us) is somehow partly constituted (or stained) by the subjective conditions of thought (e.g., our forms of intuition, our historical situation, etc.). The fact of ontological independence is then meant to establish that the known (or the in-itself) might be otherwise (or unstained). In short, it is meant to establish that how we take things to be, and how things are could be quite different. Again, the issue is that the vague account provided by Meillassoux doesn’t really spell out why there is such a dependence or what it consists in.

Now, although it is possible to give a more adequate account of thought, thereby fleshing out the initial kind of epistemic dependence posited by the circle of correlation, I think that the sufficiency thesis puts an important constraint upon such an account. The constraint is that any such account of thought must be specified in ontological terms. This is because it would otherwise be unable to leverage the ontological fact of the independence of the in-itself from thought to justify radical epistemic dependence. In essence, if the fact that there is no ontological dependence of the in-itself upon the for-us is to imply that there could be features of the for-us which are absent in the in-itself (or vice-versa), then the for-us must be understood in the same terms that the in-itself is understood in. This means that the content of thought must be understood as an entity or an aspect of entity, and that the adequacy of that content in relation to its object must be understood as a relation between entities (and their aspects). For a more detailed example of this kind of approach, you should look at my reconstruction of one of the other arguments Graham has offered for his position (here, section 2).

The important point to take away here is that, in order to shore up the sufficiency thesis, the correlationist actually needs to make a lot more assumptions than in Meillassoux’s stripped down presentation of the argument. Moreover, these assumptions cannot be purely epistemological ones, but must be grounded in some ontological interpretation of the nature of thought. This makes the whole correlationist argument a lot more controversial than it initially appears.

3. Hijacking the Argument

We’re now in a position to diagnose the way that both Meillassoux and Graham hijack the correlationist argument. In effect, they both attempt to make explicit some ontological assumptions implicit within the correlationist position. In doing this they maintain the basic relativism involved in correlationism,while locating a form of absolute ontological knowledge that this relativism presupposes. What differentiates them is the precise assumption that they take up. Meillassoux recognises the ontological character of the assumption that thought is factical, and then because correlationism still precludes knowledge of particular entities, he extends this facticity to all entities. Harman on the other hand recognises the ontological character of the epistemic dependence of thought about an object upon its subjective conditions, and then, for the same reason as Meillassoux, he extends this structure to all entities (or objects).

In each case, the fact that we cannot know anything about particular entities functions as an important premise in an argument that establishes something about entities in general. What thus unites speculative realism and OOP is that they hold that all knowledge about particular entities is relative/uncertain/impossible, but that we can thus know something absolutely about the structure of entities as such, namely, the ontological fact which underlies this relativity/uncertainty/impossibility. Meillassoux and Graham thus seem to agree with correlationism to pretty much the same extent, despite ultimately taking very divergent views. Meillassoux gives us some form of mathematico-ontology, in which the formal structure of inference (logic and mathematics) is absolutized into the structure of the world, whereas Graham gives us some form of panpsychism, in which the intentional structure of perception is universalised into the structure of entities and their relations. As an aside, there has just been some discussion of whether Graham is a panpsychist between Dominic Fox (here), Steven Shaviro (here) and Graham himself (here). I’m sticking to the guns on this one and saying that he’s a panpsychist, because I take panpsychism not to be a matter of ascribing some particular feature of thought to all entities (e.g., consciousness, will, intentionality, sensation, etc.), but simply a matter of ascribing any of them (see here). Leibniz, Nietzsche, Whitehead, Deleuze and Graham are all panpsychists, but in quite different ways.

I can now elaborate the criticism of Meillassoux and Graham’s positions I briefly sketched above. The crucial point is that the hijacking strategy only works to justify their positions if there is some independent reason to accept epistemic dependence. The core part of correlationism must be viable in order to extract some kind of absolute ontological knowledge from it. However, as we’ve seen above, more is required to make it viable than Meillassoux suggests, namely, an ontological account of thought which establishes some initial kind of epistemic dependence. On this basis I think we can see that Graham’s position is stronger than Meillassoux’s, insofar as Graham does give us such an account and Meillassoux does not. The problem is that although Graham’s account of thought does ground epistemic dependence, it can’t both ground it and be justified on the basis of it.

This touches on points I’ve made before (here and here). Graham sometimes presents his position as making consistent an independently established epistemological point about our inability to access objects in-themselves, but at other times seems to justify this point by appeal to the ontology (or metaphysics) that he derives from it. I’m not disputing that Graham’s position does make this epistemological point consistent (and thus presents something like a metaphysically consistent correlationism), but simply denying that this epistemological point has been independently established. I think many people with correlationist intuitions have been attracted to OOP, and other variants of OOO, because it reconciles two intuitions they already have: a) that our ordinary knowledge of things is always finite and inadequate, and b) that we (humans/thinkers/whatever) have no special metaphysical privilege in relation to other things. My worry is simply that if one does not already have intuition (a), or at least if one feels that this needs justification, then OOP is not going to provide it.

I should elaborate on this point a little more, to avoid possible misunderstandings of what I’m saying. Graham has sketched a number of ways of justifying the epistemological point that we cannot access the things in themselves. The most important of these is still the reconstruction of Heidegger’s phenomenological analysis of tools he provides in Tool-Being. Now, I think this is strictly false as a reading of Heidegger, but Graham maintains that we should be able to assess it on its own merits regardless of this (and he’s most certainly right). The issue is whether Graham’s particular phenomenological analysis is already tainted with certain ontological assumptions. Although I don’t want to provide an in depth reading of Tool-Being here, I think that the answer to this question is yes (see here for some of my previous thoughts on the matter). In short, I think Graham has already made some questionable assumptions (such as treating the subject, the for-us and the in-itself as distinct entities, the relations of which are then understood in metaphysical terms) in order to establish the point that things are inaccessible to us, before he universalises this to all objects. The actual universalisation itself is fairly unproblematic.

Now, I earlier hinted at the fact that there is some assumption common to Meillassoux and Graham which can be used to undermine both of their positions at once, but I haven’t as yet presented it. This assumption is a common way of understanding in-itselfness. What I mean by this is not a common understanding of what the in-itself is (on which they obviously differ), but rather what ‘in-itself’ means. There are two distinct ways to understand this:-

Mind Independence: Something is in-itself is if can exist independently of the existence of ‘minds’.

Attitude Independence: Something is in-itself if the way it is is independent of the way we take it to be.

I’d suggest that Meillassoux and Graham adopt the former, more traditional way of understanding in-itselfness, although in Graham’s case it modifies to the claim that something is in-itself if it can exist independently of its relations to other entities (insofar as all entities effectively play the role of ‘minds’). The important point is that the former is an ontological way of conceiving in-itselfness, whereas the latter is purely epistemological (insofar as it says nothing about existence). My suggestion is that it is this ontological way of understanding in-itselfness which leads Meillassoux and Graham down the correlationist path, insofar as it encourages them to understand the possibility of error (or inadequacy), and thus also the nature of thought more broadly, in ontological terms (c.f. here, section 2). The argument of the TR essay is (in part) to show how it is possible to adopt the latter way of understanding in-itselfness, which allows us to both avoid grounding epistemology in ontology, and to avoid the epistemological conclusions that lead to correlationism, speculative materialism or OOP. I won’t recapitulate this here, as it is relatively complicated.

However, it is noteworthy that this line of thought opens up the way to thinking about non-ontological dependence relations between the structure of thought and the structure of the world, and it is these relations that constitute the two new positions I mentioned earlier: deflationary realism and transcendental realism (see here). It is these kinds of dependence relations which ultimately undermine the sufficiency thesis, insofar as they allow for the possibility that the world can exist without thought, while nonetheless exhibiting a structural isomorphism with it.

4. Conclusion: Epistemology and Metaphysics

I’ll conclude by making some more general remarks about the relation between epistemology and metaphysics (and ontology more specifically). This gives me an opportunity to respond to some of the claims that Levi has been making for a while (most recently in this series of posts: here, here and here) to the effect that ontology must ground epistemology.

I can agree with one of Levi’s claims, namely, that ontology is irreducible to epistemology (contra deflationary realism and other more explicitly anti-metaphysical positions). However, this doesn’t license the claim that epistemology must be made subordinate to ontology (or metaphysics more generally). There is one very good reason for this: the question of what metaphysics is is an epistemological question.

In order to properly answer questions like ‘what are entities?’ (the question of ontology), we must be able to answer questions like ‘what is it to claim that entities are some way?’. Clearing up what it is to ask all the various specific questions belonging to metaphysics (including ‘what is essence?’, ‘what are properties?’, ‘what are relations?’ etc.) is just to pose the question of the meaning of Being, and answering them all properly is just to answer the question of Being. The important point is that to do this one must first give an account of what questioning is in general if one is to give an account of what specifically metaphysical questioning is, and that one needs to do the latter in order to do metaphysics properly. Therefore, one needs to do epistemology to do metaphysics properly. The critique of metaphysics must precede metaphysics itself.

Some might hold that there is some kind of hermeneutic circle involved here, because although one needs to do epistemology in order to do metaphysics properly, one cannot do epistemology without metaphysics. The idea would then be that we start out with some implicit metaphysical understanding which is gradually explicated and revised through the interplay between metaphysics and epistemology. I think that this is false. Regardless, the important point is that whoever wants to make this claim must demonstrate this, and they must do so in epistemological terms. Whether or not epistemology is dependent upon metaphysics is an epistemological question. This means that it is thoroughly possible to demonstrate the independence of epistemology from metaphysics from within epistemology itself. This is one of the primary aims of my project of fundamental deontology.

In short, I think the basic challenge I am putting to OOO is that although they’ve done a great deal to establish what they think about the world, I don’t think they’ve done a good enough job of explaining why they think it, and that what this reveals is a certain lack of concern with methodological issues. I could be wrong about all of the above, but I don’t think I’m wrong about their need to establish a firmer argument for the core aspect of their position. In order to do this, they need to be able to clearly say what metaphysics is in a way that is independent of their own metaphysical position (c.f. here).