Crispin Wright's "Truth and Objectivity" - Chapters 1&2

 

Introduction

I haven't written any reviews for this blog for a little while now, owing to the fact that I have a doctoral thesis to finish. This has been something I've been a little sad about, so I made a point of finding some time to write here again. I've recently finished reasonable drafts of two of my thesis papers, so gave myself permission to do a little reading, and to write a few short posts here. I decided to read Crispin Wright's "Truth and Objectivity", having had my attention drawn to the work by the upcoming Synthese special issue on alethic pluralism (a topic I write on, though from a perspective quite far from Wright's). I'm going to try to review the book in three parts. This review will cover chapters 1 and 2. Later reviews, should they get written, will cover chapters 3&4, and 5&6 respectively. However, it might be that more serious academic commitments draw my attention away from PLM reviews again, and I don't have the time to write about the rest of the book.

In this chapter Wright outlines an account of what a predicate must do in order to be a truth predicate. He then goes out to criticise two rival views (Horwich, Putnam), before outlining his own view, the superassertability view.

The ideas here are interesting for two reasons. 

First, Wright's understanding of truth leaves open a version of alethic pluralism. There might be many possible predicates that are truth predicates, perhaps with very different properties. I've been told that this doesn't go on to be a major part of the book (I have yet to finish it so won't comment). If this is the case, that's a shame! But the idea is, nevertheless, interesting.

Second, Wright's superassertability predicate is interesting in its own right as an analysis of truth. Most of this review will focus on the various candidates for truth predicates that Wright discusses, so more is said below about what I think about this predicate.

Wright frames his superassertability view as superior to the previous views of Horwich and Putnam. The impression I get, given the presentation in the text, is that he sees his view as a natural successor to those two views.

I'm unconvinced by Wright's conclusion. I think he misrepresents a certain part of Horwich's view (or, at least, is wrong about what that kind of a view is committed to), and his and Putnam's views are closer than he acknowledges.

Moreover, I think his superassertability truth predicate runs into a problem of logical scepticism. It becomes difficult to establish what the correct logic of Wright's truth predicate is.

There's an important distinction to note here between logical pluralism and logical scepticism. The logical pluralist holds that there are many correct logics and, presumably, we can know what those are be justified in using them (strictly this isn't part of the pluralist's claim, but it'll be a common feature of their view). The logical sceptic might hold that there are one or many correct logics, but that we can't know what those are, meaning that we're unable to then reason with them. I take logical pluralism to be plausible and logical scepticism to be implausible. Thus, whilst I'm excited by Wright's prospective alethic (and hence logical) pluralism, I'm worried when I argue that his view leads to logical scepticism.

As a final note, I'm aware there has been extensive discussion on this book. This is not a body of literature with which I am remotely familiar. Every point I make here might be entirely non-novel and exist elsewhere. These are just my independent thoughts on the work. Perhaps thoughts similar or more well developed exist elsewhere in print. I'd be interested to hear if they do.

Realism and Anti-Realism (§1.1, 1.2)

§1.1 outlines three ways Dummett thinks that one could be an anti-realist. §1.2 then provides a putative counter-example to Dummett's framework: humour. Humour, according to Wright, is a paradigm case of something about which we ought not be realists. Wright goes through and shows how humour cannot be captured by each of Dummett's three versions of anti-realism. Consequently, he concludes that the three conditions are not exclusive.

I won't evaluate if Wright is correct in thinking that humour can't be captured by these three types of anti-realism. Similarly, I won't outline Dummett's view. None of that is relevant to the rest of chapters 1 and 2 (though I'm informed that it is important for the later book, and Wright's motivations for writing it). None of that is also relevant for the review here.

My reply to Wright is much simpler - I don't believe we should be anti-realists about humour. Far from it, I think realism about humour should be a relatively default position, given the number of highly successful comedians who make a sustained living precicely off of the fact that their jokes or routines are really funny. It seems difficult to explain how there can be successful comedians, if we are to be anti-realists about humour.

Wright, I believe, has confused the difference between something being objective and something being real. There are many things that are real, but subjective or intersubjective, rather than objective. There really are beautiful people in the world, but who those people are depends on the individual. There really are polite and impolite actions, but what those actions are depends on one's cultural context. Objectivity is only one way something can be real.

It is true that humour is not objective. There are no propositions that glow brightly with the light of humour in Frege's abstract realm! But it does not thereby follow that humour is not real. Humour is, I suspect, a culturally relative notion. What is funny depends on what follows, given certain rules of humour (which might themselves be either culturally relative or more general), from a certain cultural common ground of shared belief, knowledge or information. Humour, then, is real, but in an intersubjective manner.

I'm therefore unmoved by Wright's objection, as I don't accept his starting premise.

This is all somewhat of an aside, though, as this content is not relevant to the rest of these chapters.

Deflationary Truth (§1.3)

The first view of truth that Wright targets is the deflationary view of truth. He attributes this view to Ramsey and Horwich.

Wright attributes the following claims to the view:

1. The Tarski Schema: T(X) iff X.
2. Truth is not a substantial property, in the sense that 1. exhausts the content of the truth predicate.
3. The purpose of truth is as a way of expressing assent (e.g. "that is true" or "everything he said is true").

From this Wright builds an argument against the deflationary view. He argues that these claims are inconsistent with each other.

Wright starts by discussing the normative role that a truth predicate must play within a conversational context. He takes truth and warrented assertability to coincide in (positive) normative force. His explication of what that means was a little unclear in places. I understand him as saying the following: Y is evidence for T(X) iff Y is a reason one may warrentedly assert that X.

This is supposed to follow from 3. but is also an independently plausible claim. It's really just a statement of the truth norm on assertability plus standard analyses of reasons, e.g. from Broome.

From this, Wright outlines the following argument. Let WA(X) mean that X is warrantedly assertable.

1. T(X) iff X [TS]
2. T(~X) iff ~X [TS]
3. T(~X) if ~T(X) [from 1,2]
4. T(X) iff WA(X) [Wright's claim iv]
5. T(~X) iff ~WA(X) [from 3,4]

5. is implausible as sometimes the evidence is neutral. In that case, X might not be warrentedly assertable, whilst not being false.

Wright takes this to be a fundamental and decisive objection to deflationism, as classically understood.

The crux, here, is point 4. the claim that truth coincides with warranted assertability. Wright is not explicit about why he takes the deflationist (even classically understood) to be committed to this. This, I believe, is a relatively unclear section of the book.

It's even more confusing when one considers that Wright has a three page footnote targeting Horwich, who he elsewhere takes to be typical of deflationists, independently. If Horwich is not supposed to be an example of a deflationist, classical understood, it's unclear who is supposed to be. I think there's a very real worry of a false-target in this argument.

As such, the following reading of Wright should be taken with a pinch of salt. Because Wright is unclear in this passage, I am uncertain of my interpretation. Wright is welcome to correct my interpretation at any point!

Wright's argument, as far as I can tell, is that if Truth is not a substantial property, but the purpose of Truth is to express assent, then Truth must be equivalent to warranted assertion (claim iv). If truth were to be anything stronger than warranted assertion, it would then be a substantial, rather than an insubstantial, notion.

The confusion seems to be about the term "substantial". This is a term Wright even admits to being unclear about in the section. Wright, I think, misunderstands what the deflationist means by "insubstantial".

Wright seems to think that the deflationist thinks Truth is insubstantial in virtue of it, effectively, being a paraphrase of warranted assertability. It is insubstantial because it makes no more demand on the world beyond the assertability of a sentence.

Here, Wright is mistaken about what the deflationist commits to. To my knowledge, no deflationist has claimed this. What I take the classical deflationist to mean, when they say that Truth is insubstantial, is that Truth is reducible to certain semantic facts, properties and relations (e.g. reference of a term, satisfaction of a predicate, etc). In that sense, Truth places no demand on the world beyond these semantic conditions holding. For an atomic sentence X, for instance, T(X) requires no more of the world than that the terms in X refer sequentially to some objects that, so ordered, satisfy the relation picked out by the predicate in X. In that sense, truth is deflationary.

This is not a new idea. This was quite explicitly outlined in Field's Tarski's Theory of Truth. It's also, I think, relatively clear in earlier work by, for instance, Tarski or Carnap.

Understood this way, however, it's clear that the deflationist is not committed to claim iv and escapes Wright's argument.

Putnam's Truth as Justification under Ideal Circumstances (§2.2, 2.3)

The next account of Truth that Wright considers is that of Putnam. Putnam claims that truth can be understood as justification under idea epistemic circumstances. For instance, an ideal epistemic agent might have seen that the sky is blue, and have no reason to doubt what they saw. They would therefore be justified in believing that the sky is blue. Consequently, it would be true that the sky is blue.

Wright raises two main arguments against this position: There might be true facts for which there could be no evidence.

The kinds of things that might be true but unprovable/undemonstrable could include certain mathematical facts (at the least, of the Gödelian kind), certain "deep" metaphysical facts, or perhaps certain ethical facts. As a relatively cogent example, suppose there was an acausal object: an object that exists but had no way of interacting causally with anything else. There could be no evidence for this object, even for an ideal agent.

A plausible response might be that Wright's reply has an impoverished notion of "ideal". Perhaps a truly ideal epistemic agent, when faced with a claim about an unknowable object, simply has a warm feeling of justification that warrants their belief. Of course, this warm feeling is reliable, by stipulation, so is sufficient for justification, given certain background views.

This wouldn't be a strong response. In explaining the powers of the ideal epistemic agent, one would likely need to refer to the concept of Truth in explaining when they have this warm feeling of justification. This would then be a circular definition of Truth. This connects to a broader worry for the view, that the very idea of ideal epistemic circumstances might presuppose a notion of truth.

A more plausible response is for Putnam to simply bite the bullet and reject that these so-say true unprovables are actually true. Motivations for this kind of move can be easily found in the movement of Logical Positivism, and the move follows from the famous Verification Principle.

Wright's response here focuses specifically on the law of excluded middle. The way he sees Putnam biting the bullet (something he doesn't see to rule out as wholly implausible, just not ideal) is by rejecting LEM in favour of an intuitionistic metaphysics.

However, the problem is broader than Wright realises. There are many logical principles at stake, not just LEM. Consider the following three statements.

1. For all statements X, an ideal epistemic agent has either a reason to believe X or a reason to believe not-X. (Wright's point, the law of excluded middle)
2. For all statements X, an ideal epistemic agent does not have reason to believe both X and not-X (law of non-contradiction)
3. For all statements X, it is determinate that an ideal epistemic agent has a reason to belief X. In other words, the agent doesn't have a reason to partially believe that X. (formally complex, depending on the degrees of partial belief available, but, typically, this relates to bivalence)

Just as Putnam cannot establish that the ideal epistemic agent has reasons for or against every proposition, he cannot establish a range of other claims, equivalent to certain logical principles in his theory.

On 2. we frequently have conflicting information such that we could rationally believe X or not-X. What reason do we have to think that an ideal epistemic situation would resolve these conflicts, without relying on a prior notion of truth?

On 3. our evidence frequently speaks in favour of certain conclusions, without being decisively in favour of it. What reason do we have to think that an ideal epistemic situation would resolve this uncertainty, without relying on a prior notion of truth?

The worry is that the only way we could come to know the truth of statements 1-3 is by becoming ideal epistemic agents, or very close to them. This is something we practically cannot do. Thus, we are forced into a kind of logical scepticism by Putnam's view. We cannot establish a great many of our logical principles from a non-ideal perspective.

A possible response here is that we can know how evidence functions and, in virtue of that, work out what an ideal epistemic situation would do, without being in one. From this, we could work out our logical principles.

This is a reasonable suggestion, however, I suspect that any plausible explanation of how evidence works, and hence what the final state of an ideal epistemic agent will be, will depend on prior assumptions about the nature of Truth. This isn't a move Putnam can make, given that he's trying to define truth.

Thus, the problem of Putnam's view is that it leads to a kind of logical scepticism that I think is relatively unappealing.

Wright's Positive Proposal: Superassertability (§2.4-§2.6)

Wright presents an interesting positive proposal that builds of Putnam's. Wright suggests that truth can be understood via superassertability (a term of Wright's own devising). The idea is an interesting one. A statement X is superassertable iff (1) a belief in X could be warranted and (2) no accumulation of additional information (or additional scrutiny of the available information) could result in the loss of warrant in a belief of X.

I understand this in the following way. Model the possible states of epistemic agents as a partial order of nodes (this is my formalisation, not his). More strictly and formally, as a meet-semilattice with a least element. This is a partially ordered structure with a least element, such that each two-element has a greatest lower bound. Informally, this is a tree-like relational structure.

Each node can be understood as endorsing a set of propositions to a certain degree. Let En(X) be true of X at a node to the extent that the node endorses X.

Intuitively, nodes are possible epistemic states of the agent.

The least node is the agent's actual epistemic state. Higher nodes indicate increased accumulation of information.

Introduce a pair of modal operators whose accessibility relation is the ≤ relation on the relational structure.

A statement is supperassertable iff it is fully endorsed at some node and then always endorsed by all higher nodes. I.e. T(x) iff ◇□ En(X) at the minimal node.

An important clarification to add is which nodes are allowed in the structure. These could be all the metaphysically possible states of the agent, holding fixed the agent's prior evidence. They could be the physically possible states (again, holding fixed prior evidence). Or perhaps some hybrid of the two.

It cannot be the metaphysically possible states. It is always metaphysically possible to discover you are being tricked in some sceptical scenario. Consequently, there would always be possible additional evidence that could undermine one's warrant and, hence, no superassertable statements. Hence the nodes in the model must be restricted to the physically possible evidential states of the agent.

It also cannot be the physically possible states. We are very constrained in our physical lives with the kind of evidence we can acquire. We cannot acquire evidence about certain facts in the distant past (e.g. was the first cave-person to start a fire right handed or left handed?). We cannot acquire evidence about the distant parts of the universe (e.g. How many stars are there in the furthest galaxy from us?). Thus if one takes only the physically possible nodes, we have a very impoverished picture.

Something hybrid is required. Holding fixed the ground-level facts of the world, one needs to consider all the states an agent could be in, given a God's-Eye view of the world. They are able to roam around everywhere, perhaps even in the distant past, and inspect things with every conceivable faculty or instrument. These are the evidential states that need to be considered in the model, if the view isn't to entail an error theory about those domains.

This is, at least on a formal level, an interesting proposal. Wright takes this to be an improvement on Putnam's view. I believe it is not a serious improvement for three reasons.

First, One boon Wright claims of his view over Putnam's is the removal of an idealised agent. There is no need to consider some possible perfect evidence gatherer, on his view, just possible epistemic extensions. I think Wright has missed the level of idealisation required in his own view. It is true that he does not require an epistemically ideal agent (i.e. a maximal node in the model) but he does require fancifully, albeit not absolutely, ideal agents. Thus, the benefit is relatively marginal, at best.

The benefit might even be non-existent. A plausible claim is that for any two epistemic states, there is a third state that is their union (i.e. a combining of the evidence of the two states). This would amount to adding a strong join-rule on the relational structure, for any set (finite or infinite) of nodes, there is a node that is their maximum. Adding this rule would entail a maximal element in the node, i.e. an epistemically ideal agent.

It's relatively easy to show how, if there's a maximal node in the structure, Wright's view is equivalent to Putnam's. ◇□  En(X) is true iff En(X) at the maximal node. Left to right: if ◇□  En(X) is true at the minimal node, then □  En(X)is true at some node, but the maximal node is seen by all nodes so En(X) is true at the maximal node. Right to left: If En(X) is true at at the maximal node, then □  En(X) is true at that node, as it only sees itself. But then ◇□  En(X) is true at the minimal node, which sees the maximal node. So Wright's view just collapses into Putnam's, if epistemic states can be combined in this way.

Thus, Wright's view is somewhere between not much better off and entirely equal to Putnam's view in this regard.

Second, the above claim about the combinability of evidence might be wrong. There are a number of cases in default logics that outline that kind of structure that justification could have whereby epistemic states might not be combinable in this way. I won't discuss those logics or these cases in detail, but the rough idea is as follows. The evidence we have effects how we should rationally interpret new evidence. Interpretation of evidence does not happen in a theoretical vacuum. Consequently, the order in which we get our evidence matters.

Consider the following case: Anna, who I know to be broadly reliable, tells me that Benjamin, who I also know to be broadly reliable, frequently lies about her. This reduces my confidence in Benjamin's testamony about Anna. I later meet Benjamin who tells me a similar thing about Anna. Arguably (I take no stance here), the fact that I already have this information from Anna, who I believed and trusted, allows me to discount Benjamin's statement. But had I got the information about Benjamin first (and responded similarly), I would have had to discount exactly the same statement that I previously accepted.

If the reading of this case is true, order of information matters and epistemic states are not always combinable.

In the model, this would mean that there are divergent branches in the partial order. Consequently, it might be the case both that ◇□ En(X) and ◇□ En(not-X), if the witness worlds for each never join. This would lead to a violation of LNC on Wright's view.

Without knowing what the more ideal nodes look like, one isn't able to rule this possibility out. Thus, Wright's view entails a kind of logical scepticism.

Lastly, my previous objection to Putnam applies also to Wright. We cannot be certain what more ideal epistemic states actually do. It's a reasonable claim that non-ideal epistemic states can (1) fail to endorse either a statement or its negation (2) endorse both a statement and it's negation and (3) only partially endorse a statement (i.e. En(X) is not fully true nor false at a node). Thus we cannot identify the logical properties of the truth predicate without better knowledge of the nature of more ideal epistemic states. Putnam's problem applies to Wright too.

As with Putnam's view, there is a great deal that could be said to defend it against the charge of logical scepticism. Largely, this would involve saying something about the nature of evidence and epistemic states. Wright must do this, though, without referencing a concept of Truth, as its Truth he's trying to define. What's more important, though, is that whatever move he can make, Putnam can too. His view will likely stand or fall with Putnam's, on this issue.

I separate the second and third objections, as the third applies to both Wright and Putnam, whereas the second applies only to Wright (the ideal epistemic agent, one presumes, gets all their information at once and hence order neturally). Thus, even if one could show that more idealised epistemic states do tend towards classical endorsements, there would still be a remaining worry for Wright in terms of identifying the correct logic for his truth predicate.

Thus, to tally the score, Wright's view is better on one point, worse on another and equal on a third to Putnam's.

Thus, I do not think Wright has demonstrated that his view is better than Putnam's.

Conclusion

 This was an interesting couple of chapters to read. I think Wright's idea of superassertability is interesting and certainly has some philosophical use, even if not in this exact context. As such, I enjoyed reading the chapters. I was largely unmoved my Wright's criticism of the deflationary view, as I take him to be mistaken about what it's committed to. I agreed with Wright's criticism of Putnam's view, in fact I think he could have gone further. However, I think his positive proposal doesn't escape the criticism of Putnam's view, and is to be evaluated on pretty similar terms.

I hope to write a little more on this book, if I have time. I'd also like to do another review of something by a younger / earlier career scholar. I've ended up reviewing more old white guys than I'd hoped! I apologise for this, as it wasn't the aim of this blog. I especially need to do a better job reviewing work by non-male philosophers of logic and maths. Suggestions are really welcome on that! I'll be doing my own research on that too.

The works mentioned in this review are:

• Wright, C (1992). Truth and Objectivity. Harvard University Press.
• Horwich, P (1990). Truth. Clarendon Press.
• Field, H (2001). Truth and the Absence of Fact. Oxford University Press.
• I defer to Wright's book for many citations of Dummett and and Putnam that I won't reproduce here.

Crispin Wright currently holds professorial positions at NYU and Stirling.

The picture is by Sergey Pesterev and is of Maligne Lake in Jasper National Park, Canada.
© Sergey Pesterev / Wikimedia Commons, https://commons.wikimedia.org/wiki/File:Sunrise_at_Maligne_lake_2.jpg
I've given up trying to find pictures relevant to the topic, so am just picking pictures I like!

I welcome comments or friendly criticism below, or you're welcome to email me at gareth.pearce@univie.ac.at

New reviews are advertised over Twitter so follow me @GarethRPearce for updates