Realism and Correlationism: Truth

This post will take a closer look at Meillassoux’s treatment of truth in Kant and correlationism. I think something crucial goes amiss here which distorts the account of so-called correlationist positions. This can make them seem vulnerable to Meillassoux’s charge that they cannot handle ‘ancestral’ events anterior to the development of life. However, this charge is misplaced, which I hope to go someway to showing in this post.

Meillassoux thinks that Kant’s “transcendental revolution” leads to a reconceptualisation of truth, since the inaccessability of the ‘in itself’ means that truth can no longer be thought of adequation with it:

From this point on, intersubjectivity, the consensus of a community, supplants the adequation between the representations of a solitary subject and the thing itself as the veritable criterion of objectivity, and of scientific objectivity more particularly. Scientific truth is no longer what conforms to an in-itself supposedly indifferent to the way in which it is given to the subject, but rather what is susceptible of being given as shared by a scientific community. (AF p.4-5)

There are a number of things wrong with this passage. For example, in describing Kant’s position, Meillassoux employs his own very simple contrast between the ‘thing itself’ and ‘for us’ in place of Kant’s own sophisticated understanding of the ‘thing-in-itself’ which is bound up with the frameworks of transcendental realism and idealism. There are more tractable problems though, which require less familiarity with Kant’s wider project in order to understand, and which have more implications for those of us who do not embrace transcendental idealism. Initially, we will need to disentangle some potential conflations operative in what Meillassoux says.

The first such conflation is between two sorts of intersubjectivity: consensus and publicity. Consensus is general agreement, whereby a group arrives at a conclusion which earns majority or unanimous endorsement. So, if we all agree that the moon is made of cheese or that Menshivism ought to be condemned, then we have reached consensus on the matter. But publicity is something very different, namely the ability to be shared — there being a common space in which multiple people, perhaps everyone, can come into a relation with something. For example, a proposition like ‘the moon is made of cheese’ is public because all suitably competent langauge users can understand it and then come to assert or deny it. So too, the moon and its properties are also public, since it exists in the same world as we all do, and we stand in all sorts of relations and networks with it — e.g. it exerts a gravitational pull upon us and our environment, emits electro-magnetic radiation which shines down on billions of people at the same time, and anyone can in principle investigate it. In contrast, candidates for privacy would be pains, illusions, dreams or phenomenal experiences (though some of these are contentious), insofar as we might think that they are only directly available to their possessors. Being private in this sense will mean failing to subsist in a common world: a territory on which multiple agents, or everyone, can interact with them equally.

The second distinction which ought to be emphasised is between objectivity and truth. Both concepts have long and contested philosophical histories, but it is only one simple contrast I want to highlight here. Objective matters are susceptible to correct responses, where ordinarily a proposition about such a matter can be true. This is sometimes captured by the label ‘cognitivism’ applied to a domain like ethics, science or aesthetics. But one can be cognitivist about a class of propositions whilst still thinking that some of them are in fact false. For instance, many people think that well-formed scientific claims are objective whilst those concerning taste are subjective, but that does not entail that all well-formed scientific claims are true. To see this, consider the claims that protons are composed of three quarks and that it is not the case that protons are composed of three quarks. Both are objective, yet only one can be true.

Meillassoux’s argument risks blurring both the public-private vs. consensual-nonconsensual distinction and the true-false vs. objective-subjective distinction. If so, this would be fatal. The most charitable interpretation of the argument he attributes to the correlationist which I can construct is the following:

(1) If things-in-themselves are inaccessible, then objectivity cannot be conformity of representations to things-in-themselves. (Premise)

(2) Things-in-themselves are inaccessible. (Premise. Established by the ‘Short Argument’)

(3) Therefore, objectivity cannot be conformity to things-in-themselves.

(4) Either objectivity is conformity of representations to things-in-themselves or it is a property of subjective representations. (Premise)

(5) So, objectivity is a property of subjective representations.

(6) Objectivity of representations requires universalisability of representations. (Premise)

(7) Universalisability of representations requires their capacity to be shared [perhaps by everyone] (Premise)

(8) Therefore, objectivity of representations requires their capacity to be shared.

(9) Scientific truth requires objectivity of representations. (Premise)

(10) Therefore, scientific truth requires representations to conform to the conditions for being shared.

On this charitable reading, the argument begins by trying to show that objectivity must be a property of representations rather than a relation between representations and things-in-themselves. Here, I think the most plausible way to understand objectivity is as a kind of semantic or epistemological form, which supports a distinction between something being the case and it merely seeming to someone to be the case. Meillassoux would thus be saying that both pre-Critical and post-Critical philosophy can understand objectivity as the possibility of making a distinction between being and seeming to be. But where pre-Critical philosophy would frame this in terms of truth as adequation (or conformity) of representations to things-in-themselves, this option is supposed not to be open to the correlationist since for them we alledgedly cannot represent things-in-themselves. So, the correlationist needs a new way of thinking about objectivity, which does not breach the circle of representations. This they find in the idea of universalisability: if a representation can be universalised, and thereby could be possessed by everyone, it is objective. If the representation is not available to everyone in this way, it is merely subjective, and can only count as a representation of how thing seem to its possessor. Since scientific truth requires objectivity, then the correlationist is supposed to be committed to scientific truths being conditional upon representations being capable of being shared. Thus, scientific truths, within a correlationist framework, are anchored to conditions of shareability, and are unintelligible without them. From here, it is not far to the claim that correlationism cannot cope with putative truths about events anterior to life, where such conditions could not obtain.

This version of the argument is vulnerable on a number of counts. Most of the premises are controversial, especially if they are understood in the senses required for the argument to be successful. So too, even though it is intended as a reconstruction of a Kantian line of thought, it does not map onto Kant’s actual claims. But things are even worse for Meillassoux, since I think his presentation of the correlationist argument is even more flawed than this version here. We can start with what he explicitly says, and then work back to what I take to be the stronger version of the argument as presented here.

In presenting the argument, I have sought to mark the distinctions which I outlined earlier. But Meillassoux seems guilty of blurring them in a problematic fashion. For example, on behalf of the Kantian correlationist, he says:

The difference between an objective representation (such as ‘the sun heats the stone’) and a ‘merely subjective’ representation (such as ‘the room seems warm to me’) is therefore a function of the difference between two types of subjective representation: those that can be universalized, and are thus capable of being experienced by everyone, and hence ‘scientific’, and those that cannot be universalized, and hence cannot belong to scientific discourse. From this point on, intersubjectivity, the consensus of a community, supplants the adequation between the representations of a solitary subject and the thing itself as the veritable criterion of objectivity, and of scientific objectivity more particularly. (AF p.4)

But this conflates publicity and consensus. Without a conception of truth as adequation to draw upon, we might think that the capacity of representations to be shared ought to count as the criterion of objectivity. But there is no reason why this should lead to consensus of communities being invoked alongside it. That anyone can share a representation might be thought to bolster its status as a glimpse of the world — even if it must be the world ‘for us’ — since trans-subjective rational constraints will be operative upon people’s cognition: there will be some sort of common space which overspills each individual agent’s world-view. This would allow a convergence of people’s judgements, but such consensus will not be a condition or criterion of objectivity, since it can be present when there is no objectivity (people may just happen to agree) and it can be absent when there is objectivity (some people may just be wrong). Thus, consensus seems besides the point, and it distorts the line of thought Meillassoux attributes to the correlationist. So, in my presentation of the argument, only publicity (i.e. shareability) is invoked.

Secondly, Meillassoux risks running objectivity and truth together. For the correlationist, he says, “Scientific truth is no longer what conforms to an in-itself supposedly indifferent to the way in which it is given to the subject, but rather what is susceptible of being given as shared by a scientific community.” (AF p.4-5) But this last clause is how he describes objectivity; and it seems plain wrong to say that scientific truth is ‘what is susceptible of being given as shared by a scientific community’. The best sense I can make of this is to suppose that he means to say that the conditions for being given as shared are conditions for objectivity and objectivity is a condition for scientific truth. I try to reflect this in my presentation of the argument too.

Even with these revisions in place, it seems to me that Meillassoux mischaracterises the thrust of the Kantian strategy. Kant is not trying to redefine truth or objectivity in intersubjective terms, under the pressure of epistemological constraints introduced by transcendental idealism. Instead, he attempts to vindicate certain a priori concepts — such as the categories of the understanding — as being objectively valid. For example, these concepts include like causality, as a necessary connection between two events. These concepts figure in Kant’s attempt to provide a reformed and legitimate metaphysics, able to justify the concepts to which it appeals. In contrast with empirical concepts, such as bear or atom, we supposedly cannot give a full defence of them by simply looking to the world and seeing whether there is anything which corresponds to them (recall Hume’s scepticism about justifying causality). For Kant, these concepts have a special status: “since they speak of objects through predicates not of intuition and sensibility but of pure a priori thought, they relate to objects universally, that is, apart from all conditions of sensibility.” (B120) Not being based upon experience, they “arouse suspicion.”

New strategies of justification are thus required, where Kant attempts to undertake transcendental deductions of a priori concepts. The most famous of these, in the first Critique, tries to justify categories of the understanding, and has two parts. The first tries to show that these categories are conditions of thought which are necessary (no cognition is possible without them) and universal (they are conditions upon all cognisers). The second part tries to show that these are not just subjective conditions upon cognition — perhaps peculiar to human biology and how we happen to have to think — but equally conditions upon objects, such that objects themselves must conform to them (e.g. objects must be in a causal order, be possible, actual or necessary, etc.).

So, the role in which Kantian appeals to universality (and publicity) appear are not as new criterions for objectivity or truth. Rather, universality features only as a first step in an attempt to justify a special set of a priori concepts, and even then it is nothing like sufficient to show that these concepts are objectively valid. Kant’s conception of truth remains a fairly standard one: “Truth and error […] are only to be found in the judgment, i.e. only in the relation of the object to our understanding.” (A293=B350) His account of concepts as rules means that his understanding of truth is probably not best captured by the idea of ‘adequation’ (e.g. it does not rely upon a sort of picture theory of meaning). Nevertheless, it is important to stress that the label ‘intersubjective’ is even more misleading.

There are further issues here. For, we might think that Kant’s whole epistemological framework is subjectivist. It is all very well to say that truth is a relation between objects and our understanding, but if objects are mere representations, or constituted by subjects, or somesuch, then this talk is cheap. If we approach Kant in this spirit, then few of the above considerations will move us. Even amongst those who would never confuse transcendental idealism with the material or methodological idealisms of Berkeley and Descartes, this approach remains. It certainly seems to underlie Meillassoux’s concerns. But whilst I think that Kant does not quite present us with a sufficiently desubjectivised epistemological framework, his position is far removed from this picture, both in its aspirations and its salvagable achievements. But a defence of this conviction would be a massive undertaking itself. I have little to say about it here, except to point to the fruitfulness of recent scholarship in partnership with frontline non-historical work which it has informed. To point to just one issue, promising ways of understanding the phenomenal-noumenal distinction and the limits on knowledge signalled by the concept of the in-itself have been established, without sliding into scepticism or ontological dualisms. Despite their disagreements, the work of Sebastian Gardner, Fred Beiser, Graham Bird and Henry Allison (as well as post-Woodbridge McDowell), helps us see what this sort of Kant might look like.