Brute Counterfactuals and Causation

Truthmaker principles were first used by [inter alia] Armstrong to object to phenomenalism about objects and behaviourism about the mental. These theories both analyse statements about objects or about mental properties into a bunch of counterfactual statements about what would occur or be observed under various conditions. Since those counterfactuals themselves are left unanalysed, the truthmaker principle deems them illegitimate: since every truth requires a truthmaker, and these theories require brute counterfacutals, so much the worse for them. Why is this sort of objection any weaker against counterfactual analyses of causation? I find this quite compelling: surely we still need an account of why it is the case that if C hadn't occured, E wouldn't have?

Absences, Causes and Explanations

There's prima facia compelling evidence that absences are causally efficacious. We are happy to say things like: My forgetting to water the plants caused them to die; the fire was caused by my not blowing the candles out. I think these statements are strictly false, that this shows that locutions of the form 'a caused b' are ambiguous, and that this has helpful implications for the take on truthmaker theory outlined in the previous post.

If we take the linguistic evidence for the causal efficacy of absences at face value, then we must regard causal determination as depending heavily on human interests. We regard the absence of my watering the plants, but not the absence of someone else on the other side of the world watering the plants, a cause of their death. The death of the plants counterfacutally depends on both absences: if I, or anyone in China, had watered them, they would have lived. The difference lies in the fact that my failure is salient, but theirs is not.

But we should not think of causal relationships determined by what is salient to us. We should the world as causally structured independent of our concerns, a structure it is the business of physicists and chemists to describe.

Given the fact that we should regard causal relationships as objective, and that we are happy to call causes some but not other abences, it follows we should not regard statements of the form "the absence of A caused B" as truly asserting that the absence of A caused B. Rather, we should read such statements as asserting that the absence - causally - explains B. [Two ways to go: "A caused B" is ambiguous and sometimes literally means "A explains B"; or sometimes "A caused B" is literally false but implicates truly that A explains B. I don't know why to prefer one to the other.]

According to Lewis, to explain an event is to give information about its causal history. So one can give information by describing the absence of a certain type of cause. No problem here. Here's a neat explanation, moreover, of why we use "cause" locutions to make claims about explanation: statements relating facts are naturally formed using "because", which is neutral between causation and explanation, while there is no natural way to form statements relating events without using "cause".

The proposal considered in the previous post was to replace the truthmaker principle:
[TM] For every truth, there is some object in the world that makes it true
with
[Exp] For every truth, there is some [ontological] explanation of its truth.
The problem of finding truthmakers for negative existential truths bedevils truthmaker theory; we can now see that [Exp] avoids this problem comfortably. For the explanation of the truth of "there are no unicorns" is precisely the absence of unicorns.

Postscript: Another virtue of this account: once we make a distinction between uses of "A caused B" to make explanatory or literally causal claims, we can defend the transitivity of causation too. Counterexamples look like this: a train hurtles toward a cow on the track. I pull a lever making the train switch tracks, thinking I have saved the cow. Unfortunately, the tacks merge shortly before the cow. I caused the train to switch tracks. This, in turn, caused the train to rejoin the original track. This caused the death of the cow. But I did not, we protest, cause the death of the cow. [A caused B, B caused C, but not[A caused C]] The solution: I did cause the cow's death, but we balk at saying so since my actions do not figure in any appropriate explanation of the cow's death. [I'll need to think about whether "a causes b" always suggests the explanatory claim and always suggests the causal claim.]

Truthmaking and Explanation (Updated)

One charge against truthmaker principles(made by e.g. Merricks in Truth and Ontology) is that far from being providing a way to decide between metaphysical positions, they are vacuous unless we make substantial metaphysical assumptions. I have a suspicion, however, (a rather hazy one) that we can save truthmaker theories from emptiness if we pay attention to issues concerning explanation. (By the way, I'm getting so obsessed that I'm now tempted to think every philosophical problem is really about explanation.)

One motivation behind truthmaker theory is that it provides a way to make substantive the thought that truth depends on being: that things are true in virtue of the way the world is; that what there is and how it is explains why true things are true. Consider one version of the principle (Truthmaker Maximalism + Necessitarianism):

(TM) For all truthbearers P, P is true -> there exists some m such that necessarily, [(m exists) -> P]

But take an arbitrary truth, P1. Then the state of affairs, m1:that P1 is such that necessarily, if m1 exists, then P1. In particular, substitute for P dodgy counterfactuals or present Lucretian truths about the universe's history for presentists, which are supposed to offend TM, and we can still satisfy TM:

P1: Archimedes was killed in a battleM1: the state of affairs of the universe being such that Archimedes died in a battle.

The only response left for the truthmaker theorist seems to be to exclude these moves on the basis that they involve "suspicious" states of affairs or properties. But that just tells us that it is not TM but our other substantial metaphysical commitments that inform us what makes truths true.

Here's my remedial thought. Truthmaker theory is essentially an explanatory enterprise. (Although truthmaking is stronger than explanation - my giving you the money explains why you bought the icecream but is not a truthmaker for it.) As the familiar objections to the Deductive-Nomological model of causation show, there is an explanatory direction to the world: in general, if P explains Q, Q doesn't explain P. And for no P does P explain P. The same will hold for groundings. Higher-order phenomena are grounded by lower-level phenomena. Facts about tables are grounded in, made true by, facts about their constituent particles etc..

The upshot is that one has failed in the explanatory enterprise, for example, if one cites a bare counterfactual as part of a truthmaker for a counterfactual. Of course, some facts are brute, and I just assumed counterfactuals are not. When we have disagreements about which facts are brute, then TM is genuinely useless. It is promising that the same considerations will exclude TM1: if anything, the P1 and TM1 have an explanation in common; one cannot see TM1 as an explanation for something without also for the same reason seeing P1 as an explanation on similar lines. So to cite TM1 as an explanation for P1 is again to fail the explanatory enterprise.

Two exciting conclusions:
(1) since we can often agree about which facts are brute, TM will be useful in many metaphysical debates.
(2) There is important philosophical work to be done figuring out on principled grounds which facts are brute. [Cue jokes about G. W. Bush being a brute fact...]

An objection:
(1) We all thought that a truthmaker for "Fido is on the mat" is the state of affairs of Fido sitting on the mat. But since this fact is not brute, to cite it as a truthmaker is to fail the explanatory enterprise. I think we should bite this bullet.

Update: some more implications to think about(1) What sense of explanation is at work? Plausibly some sort of determination relationship, but it is commonly agreed necessitation is too weak. (My thumb necessitates, but is not a truthmaker for, 2+2=4)
2)We saw the truthmaking relation (the relation that obtains between a truth and its truthmakers) is explanation plus something else. A plausible thought is that truths are true in virtue of their truthmakers’ existing. How to spell this out?(3) Explanation is context-sensitive: what counts as a good explanation of a fact depends on the context of assessment. Does this context-sensitivity carry over to truthmakers?

Peter Lipton RIP

I have just learned from Leiter Reports that Peter Lipton died of a heart attack on 25th November while playing squash. The Cambridge Department of History and Philosophy of Science(HPS) has a short obituary here.

Although I did not know him personally, I am deeply saddened by the loss to the philosophy profession of a first-rate philosopher. Peter Lipton is the author of the only book-length study of inference to the best explanation, and I have learned from his work tremendously, not only on IBE but in the philosophy of science more generally.

I've thrown myself at tackling the problems Lipton's defense of IBE faces, and there lies an enormous amount still to be thought through. I'm convinced that a plausible account of IBE lies at the heart of the solution to a lot of philosophical problems, and while Peter Lipton has made an admirable contribution to that aim, it is a great shame he will not be able to guide the debate in the future.

Return to Blogging!

I haven't blogged for a long time, and I've missed being confused about such desultary issues that online navel-gazing allows. So I have promised myself to get thoroughly perplexed over the next few days engaging with the following:

(1) The Truthmaker principle (TM) says that for every truth, there exists something that makes it true. But this is just too easy to satisfy. You say:

(P) If Marco had been born in ancient greece, he would have been Aristotle's slave.

I object: (P) has no truthmaker. You retort: the truthmaker for (P) is Marco's being such that if he had been born in ancient Greece he would have been Aristotle's slave. Any demand for a truthmaker can be met with such ad hoc truthmakers. So if TM is to be a substantial thesis, we need a principled way to rule out those 'suspicious' properties. Since TM was meant to be a way of deciding between metaphysical positions, some take the problem of easy truthmakers to be fatal.

But look: the motivation behind TM is that 'truth depends on being' in some substantive way; that what there is and how it is explains what is true. I'm trying to spell out the idea that the
same explanatory considerations that motivate TM will rule out suspicious truthmakers.

(2) I've thought a fair bit about IBE since the last post, although not too much recently. I'll post about why charges of damaging context-sentivity to IBE fail

(3) Duhem was a fin de siecle French (see what I did there?) scientist to whom both instrumentalism and anti-explanationism are attributed. (By, for example, van Fraassen and Cartwright.) I think he held neither of these, and in fact to read these views into him makes him state blatant contraditions. His position is pretty sophisticated - so it's difficult to find a position that would make him consistent! - but the best one we have in some form of restricted Structural Realism.

(4) Frege's notion of sense was meant to perform a variety of roles. Importantly, it both determines the reference (say, of a singular term) and accounts for the 'cognitive significance' of identity statements involving that term. I want to argue that the defense of such a notion essentially involves appeal to the principle that the content of our beliefs is explanatory.

(5) I went on the the St Andrews Philosophers' trip to Raasey last month. There were some really enjoyable papers given there. Mark Wales' in particular was very fun: he tried to convince us of the merits of an account of knowledge that doesn't require truth. As I said, entertaining. But I thought It'd be important to make some of the formidable objections clear.

Inference to the Best Explanation

It turns out to take some thought to flesh out the slogan of the principle of inference to the best explanation. And it turns out, surprisingly (to me at least), that how to flesh it out is controversial.

Lipton (1991) gives an account that sounds plausible and that has attracted many supporters: inference to the best explanation is inference to the hypothesis that, if true, would explain the best (which he takes = confers most understanding). The problem of defending IBE then looks like it has to take the form of arguing that hypotheses that explain the best are more likely to be true.

But it doesn't take much thought to realise that what explains the best and what is likely to be true often rub in opposite directions. Consider the conspiracy theorist: he has a very detailed explanation that, if true, would allow her to explain many apparently unrelated events. The theory is, of course, wildly implausible.

So here's an account that might do better: inference to the best explanation is a principle in two parts. The account also has the benefit of being directed on a particular fact to be explained, which better describes how appeals to IBE are in fact made. So, take some phenomenon to be explained, say the repeatability and structure of our experience (E). The principle of IBE states:

[i.] Consider the set of all the possible explanations for E. (Including (H1) the evil demon hypothesis and (H2) the external world hypothesis.) One of these must be true: every particular fact has an explanation.

[ii.] Occam's razor: simpler theories are to be preferred over their rivals.

Assume for the sake of the argument that H2, the external world hypothesis is simpler. (See Vogel (1991) for an argument to this effect.) From [ii.], it is to be preferred over H1. This is not quite enough, since better needn't imply good: b1 = I will win the lottery this week is better than b2= I will win it this week and next, but it is, unfortunately, no good! Even if we knew that one of a set has to be true, that needn't make one justified: take b1 = I, having two tickets, will win the lottery this week; b2= bob, having one ticket, will win the lottery this week; b3 = bill, who has one ticket, will win this week ... and so on for everyone in the lottery. One must be true, and b1 is more likely to be true than any other. But b1 isn't justified.

We might blame it on the fact that b1 is less justified than b*= b2 or b3 or b4 or ... . What if we take the competing beliefs more generally so that can assert the disjunction: either the explanation for the nature of our experience is that there is a world of objects like chairs and tables that causally interact with each other and causally impact on observers, or we are fed experience that is as if this occurs. We would need to argue that even in the most general form, what is shared by any theory of this type is less simple than the external world hypothesis. So what we need to defend is:

1) If we justifiably believe that a disjunction (b1 or b2) is true, and know that one, say b1, is preferable = better justified, this justifies believing b1

(2) That in general we can formulate inferences to the best explanation as just involving two general hypotheses such that all particular explanation fall into two types, and that we can see one class as sharing a simpler structure than the other.

Synthetic Apriori True Justified Belief

Arguably, it is possible to learn about which metaphysical theses are justified and which aren't. Say, whether we shold think that particulars are bundles of tropes or bare things which instantiate universals. Arguably, there is a truth of the matter: a theory that reflects how things in the world stand. But if we hit on the correct metaphysical account, and we have good reasons for holding to our account, we have justified true belief in metaphysics.

So metaphysics involves aiming for synthetic apriori knowledge, as long as we're willing to be fallibilists about knowledge.

Real-life dispute about Explanation!

Mind states correlate with brain states. This much is widely accepted. It is consistent with a range of dualist positions, but can we argue that this fact favours an identity theory on explanatory grounds?

In Hill, Sensations (1991), and McLaughlin, "In Defense of New Wave Materialism" (2001), both present an argument whose core structure is something like that presented below.

1. We have the following phenomenon to be explained (It's not clear to me that any potency in this argument depends on talking of type rather than token correlation and and identity) :

(Correlation) For every type of mental state M, there is a type of physical P state such that it is nomologically necessary that something is in M iff it is in P.

2. Now consider the following thesis:

(Identity) For every type of mental state M there is a type of physical state P such that M=P.

While (Correlation) is consistent with interactionism and epiphenomenalism as well as (Identity), the latter thesis provides the best explanation for (Correlation).

3. If a theory provides a good explanation for a set of facts F, and this explanation is better than any explanation provided by a competing theory, there is good and sufficient reason for believing that the theory is true.

4. So, (2.) is a reason to think (Identity) is true.

Kim, in Physicalism, or Something Near Enough makes some interesting objections:

(i) Identities can't explain anything. In science, correlations are explained by appealling to a lower-level structure or process underlying the correlation. Once one has an explanation, identities can be used to rewrite the explanandum: we have an explanation of x, x=y, so we have an explanation of y. This schema doesn't give us a new explanation, just a new way of putting an existing explanation.

(ii) Identities rule-out correlations, and therefore requests for explanation. As in Smart (1959): 'to say that they are correlated is to say that they are something "over and above."'. If mind states merely correlated with brain states, we should be enbarrassed for the lack of an explanation of that fact; that fact that we have (identity) frees us from the obligation to have an explanation. Similarly, asking for an explanation of why heat correlates with mean molecular kinetic energy is just sidestepped when we know that an identity holds between them.

Kim also points out that, given (ii), although it might be true that we are better off from a explanatory piont of view with (Identity), this is no longer subsumable under a principle of inference of best explanation. We merely have a reason to want to think (identity) is true, and do not actually have one.

Contra Kim, Block and Stalnaker ("Conceptual Analysis, Dualism and the Explanatory Gap") hold that theoretical identities are justified on explanatory grounds: If we accept such identities as water = h2o, heating water = increasing mke of h2o molecules, boiling of water = certain molecular activity of h20 molecules, we have an explanation of why heating water causes it to boil. If we replaced the identities with correlation, all we'd have is an explanation of why something correlated with heating causes something correlated with boiling. So believing the identities allows us to explain facts we otherwise couldn't. So we are justified by the principle of inference to the best explanation.

Kim's retort is to point out that the explanatory benefit acheived by the identities, although real, is not warranted by the principle of inference to the best explanation. Identities allow explanations to be rewritten, but do not provide new explanations.

So we can distinguish two sorts of appeal to inference to the best explanation that uses what Kim (in "Explanatory Realism, Causal Realism and Explanatory Exclusion) has called an 'objective explanatory relation'. (Causation is an obvious one, but is seems quite possible that there be others: supervenience, "the micro-reductive relation")

a) If E is the best explanation of F because it cites only objective explanation relations between E and F, there is reason to believe E.

b) If E is the best explanation of F, although it appeals to facts that aren't objective explanation relations, there is reason to believe E.

The charge is that a) is a valid form of inference, b) isn't. What I'm interested in is how we can exclude some facts (like identities) from counting as an "objective explanation relation", and how we decide what is to be included.