Wednesday, August 03, 2005

Experiencing Modalities

Why do many philosophers seem to have accepted without argument that experience can only give us epistemic access to how things are, not to how things must be or how they could be? As far as I can tell, Plato, Kant and Hume were all convinced of the truth of (some version of) this claim, and it also has later advocates (I can think of relevant passages in Whewell and Chisholm).

But even if we thought that all modal truths were knowable a priori (and we had not yet appreciated that a priori knowledge may be empirically grounded), what would be the grounds for denying that experience can provide some (a posteriori) form of epistemic access to modal truths?

Granted, we never see modalities or bump into them. But we never see or bump into quarks or dark matter or magnetic fields either. We (or rather, those experts to whom we delegate such things) infer to their existence as the best explanation of what is observed. This is a perfectly respectable way for experience to provide epistemic access to some swathe of reality.

Is it perhaps assumed that modal truths cannot serve as explanations of the things we observe? Or is it that some other, non-modal explanation will always be better?

Alternatively, is the assumption that experience cannot provide us with epistemic access to modal truths just outdated? Has our epistemology moved on, with the realization that we can have empirical grounding for believing in something without actually bumping into it, and left this old-fashioned assumption behind?

27 comments:

Daniel Elstein said...

I agree with you that modal truths are involved in explanation, but I don't think they're explanatory in the normal way, so the analogy with quarks seems a little quick. It looks plausible that explanation is itself a modal concept. The connection between causal explanation and counterfactuals is pretty standard: if I say that the presence of an object explains an observation, I'd better also be prepared to claim that a different observation would have been made if that object had been absent (ceteris paribus) on pain of giving a bad explanation. This goes for other kinds of explanation too: what makes the fact that the room is 10' by 20' a good explanation of why you need 200 sq feet of lino to cover it is that the explanans necessitates the explanandum. Since acquiring knowledge through inference to the best explanation depends on the explanations being (known to be) good, there's no such inference without modal knowledge. You can't know that there are quarks without knowing various counterfactuals which make the existence of quarks explanatory of observation. And knowledge of counterfactuals about quarks has got to be a posteriori at least in part, because you've got to know about quarks before you can know what counterfactually depends on them (it's not like knowing that e.g. each quark is self-identical).

But, and it's quite a chunky but, it's not as if the modal truths themselves are part of the explanation. It's not the modal truths about quarks which explain stuff, it's the quarks. So there's something very fishy about seeing the indirect knowledge of modal truths as being like indirect knowledge of quarks. It tempts you into thinking that there are non-actual objects which do the explaining. And it also makes you think that there's some special kind of modal explanation, to be compared with e.g. causal explanation. That's when the a posteriori route to modal knowledge starts to look suspicious, since it is mysterious how we could know about non-actual talking donkies via observation. After all, they aren't causally connected to us, and we want to know what other kind of connection there is - modal connection?

The upshot is that scepticism about modal a posteriori knowledge is wrong, but it isn't attributable to failure to realise 'that we can have empirical grounding for believing in something without actually bumping into it'. Rather, it's the failure to realise that knowing about modality is the way we do our abuction ('the way we abduct'?), and it doesn't involve knowing about some mysterious unconnected objects. More succinctly, the error is modal realism, or modal descriptivism, not naive epistemology.

Carrie Jenkins said...

Hi Daniel,

I'm thinking that my main opponent here isn't a modal sceptic but rather someone who thinks modal knowledge is only obtainable by non-empirical means.

If you're thinking that what you described is the *only* way that modal truths get involved in explanations, you seem to be on their side, since then you've ruled out both direct and abductive empirical knowledge of modal truths. (I take it that direct knowledge is obviously implausible, and modal knowledge couldn't be abductive if this view is right, because it's actually prior to abduction - it needs to be in place before we can do any inferring to the best explanation.)

But what are the reasons for thinking that this *is* the only way modality gets involved in explanation? You're suspicious of quark-like kinds of involvement because you think 'it is mysterious how we could know about non-actual talking donkies via observation. After all, they aren't causally connected to us, and we want to know what other kind of connection there is - modal connection?'. But it's not obvious to me that inferring to p as the best explanation of what we observe requires that there be 'a connection' between us and p (other than the explanatory connection between p and what we observe).

Daniel Elstein said...

Well, I do think that the way I've described is the only way. I'd want to say that whenever a modal truth looks explanatory, it's standing in for something non-modal. I'm not sure whether I want to try to *argue* that it is the only way. We have an intuition that modal truths are explanatory because they're involved in explanations. I want to explain that intuition away by showing how modality and explanation are inter-related. But if the intuition remains in place, presumably it's grounded on some cases where there seems to be a different kind of explanation. So I guess I'm committed to answering those cases when they come up, but I'm not sure how to rule them out in advance. Did you have any examples in mind?

As to whether I'm excluding a posteriori modal knowledge; well, it would be bad news if I were. In order to know whether the tumbler would break if it were dropped, I have to know what the tumbler is made of, and I know that a posteriori. So my knowledge of the counterfactual had better also be a posteriori. I don't agree that on my account 'modal knowledge couldn't be abductive, because it's actually prior to abduction', though I see that what I said looks very close to that. Trying to be more careful, I'll say say how I think I can avoid that conclusion. And I'm afraid I'll say lots of things that I haven't though through properly...

When we give a causal explanation, we suggest the presence of objects with certain dispositional properties. Dispositions entail counterfactuals analytically (or conceptually, or whatever). What makes the explanation good is that those counterfactuals are the ones you want. For example (very roughly), in explaining why the glass broke a good partial explanation is that it is fragile; a bad explanation is that it's transparent. The counterfactuals generated by transparency are clearly all wrong when you're interested in why the glass broke. Abduction involves comparing good explanations and somehow choosing the best. Once you've chosen the best explanation, supposedly you know it to be correct. So you know know that the objects with the relevant dispositions are in place, and thus you know that the counterfactuals which are entailed by them are true. So you know modal truths by abduction.

I should say that the position taken here on dispositions is very crude and probably wrong. You could easily just say that knowledge of dispositions just is modal knowledge, and that when an explanation involves dispositions it's really the modal stuff that's explanatory. My vague intuition is that it's objects with dispositional properties that are explanatory, and that it's dispositional properties all the way down, at least as far as we can tell. Sorry for the long answer, I really should work out what I think about this stuff one day.

Brad Majors said...

I take it that the basic reason experience cannot provide knowledge of how things must be is that our experience gives us access only to the actual world. And modal truths -- at least those involving necessity -- concern all possible worlds. You're certainly correct that unobservability isn't sufficient for lack of empirical access. But facts about quarks and dark matter are still contingent (presumably). And I see the main issue here as revolving around the relation between experience and necessity.

Of course, not all necessary truths are knowable a priori. But they arguably all involve an a priori knowable component (the necessity of identity for 'Hesperus=Phospherus', the necessary or individuative nature of microphysical structure for natural kind terms in the case of 'Water=H2O', etc.)

And not all truths knowable independently of experience are necessary. But for every case of the contingent a priori, I take it, there must be some explanation for how the relevant content can be known true in the absence of knowledge of the nature of the actual world. For the cogito (as well as related judgments such as 'If I exist, and this place here exist, then I am here') this guarantee is provided by the nature of judgment. For Kripke's example of the metre bar, the guarantee is provided by the phenomenon of reference fixing. And so on.

Why, then, cannot there be *some* a posteriori access to modal truths? Because, in the general case, modal truths depend on how things are in all possible worlds, and experience can tell us only how things are in this world. (Two-dimensionalism complicates this, but I think that the central point remains.)

There is nothing new here, of course. I'm just trying to articulate the standard rationale for the view you are questioning, with a view toward continuing the dialogue.

Carrie Jenkins said...

Daniel,

I'm not sure I'm happy about thinking of dispositional properties as non-modal, though it's an interesting idea: let me encourage you to continue thinking about it! But counterfactuals aside, would you say that knowledge of straightfoward claims of necessity and possibility can never be empirical?

"Did you have any examples in mind?"

Not really, but let's try a simplistic one (that should be an easy target) to start off with. Suppose I say the best explanation of why all water is H2O is that water is necessarily H2O. What sort of mistake have I made?


Brad,

Hi and thanks for the comment.

You write:
"Why, then, cannot there be *some* a posteriori access to modal truths? Because, in the general case, modal truths depend on how things are in all possible worlds, and experience can tell us only how things are in this world."

But I'm worried that what comes after the 'because' here looks to me a bit too much like a restatement of the thing it's supposed to explain. Unless I'm already convinced that experience cannot deliver modal knowledge, why should I accept that "experience can tell us only how things are in this world"?

Matt Kuenning said...

Maybe Carrie's questions to Daniel and Brad can be answered by Wright's notion of the failure of warrant to transmit across certain known entailments. Imagine some p that I know empirically. I also know that if p, then possibily p. But it seems I don't know possibly p empirically, because if I didn't know (or weren't entitled to) the proposition that possibly p, then whatever empirical evidence I had for p wouldn't have been evidence for p.

One virtue of the above idea is that it accounts for the strong intuition that no experience can be evidence for propositions that are not possible - i.e. those whose negation is necessarily true. I see something that looks round, now square, or I measure the interior angles of an apparent triangle and get 178 degrees. Whatever this is evidence for, it isn't evidence for round squares and 178 degree triangles because no one is entitled to the proposition that such things are possible.

I think Carrie's question about water is amenable to this sort of answer. If you weren't entitled to the proposition that water is necessarily identical with its micro structure or chemical composition (because it's a natural kind or whatever), then your empirical evidence wouldn't be evidence that all water is h2o. Consider that the term "soil" is unlike "water" in that respect. That's why, when a soil analyst gets evidence that the soil in this field is xyz, its not evidencee that all soil is xyz. If so, then even if the best explanation for the fact that all water is h2o is that it's necessarily so, the empirical warrant for the proposition about all water doesn't transmit across the explanatory link.

I gather Daniel may have been driving at the same point by saying that modal truths are involved in or presupposed by explanations but are not properly part of the explanation.

Carrie Jenkins said...

Hi Matt,

Interesting stuff ... is the idea that the *only* way experience can provide evidence for possibly p is through providing evidence for p, from which we then deduce possibly p? If it's not the only way, then even if there is transmission of warrant failure in these particular cases, that wouldn't rule out empirical knowledge of possibly p through some other route.

I'm not quite sure I'm seeing the case for transmission of warrant failure here. You write:
"if I didn't know (or weren't entitled to) the proposition that possibly p, then whatever empirical evidence I had for p wouldn't have been evidence for p."
But why? You mention an intuition that no experience can be evidence for something which is impossible; is that supposed to do some motivating work here? (If so, we need to be a bit careful about whether the restriction is to propositions which are possible or propositions which are known or justifiedly believed to be possible.)

I don't actually have the intuition that no experience can provide evidence for something impossible (although I know that others share it - if I remember rightly, Eliot Sober has defended something like this, and he goes on to argue that because of this nothing can be evidence for a necessary proposition either). Here's another simple case: suppose I hear a well respected mathematician earnestly asserting some mathematical proposition which is (unbeknownst to her or to me) in fact false. Doesn't what I hear provide me with evidence for the truth of the proposition, even though it is impossible for it to be true?

It might be more plausible that no experience should be taken as evidence for something that is *known* or *believed* to be impossible (though even here I have my doubts). But if this is all we have the transmission of warrant worry doesn't seem to be motivated. Suppose I don't have any prior opinion as to whether p is possible, but then acquire empirical evidence that it is true and deduce that p is possible. You'd want to say I relied on the possibility of p in accepting my experience as evidence for p. But that claim can't be motivated by the thought that we shouldn't take something to be evidence for p if we believe p is impossible. So what motivates it?

Maybe the motivating thought is that we shouldn't take something as evidence for p unless we *already positively know* (or justifiedly believe, or are entitled to accept, or whatever) that p is possible. But that doesn't seem right - surely evidence for p and for possibly p could arrive simultaneously (as in the case of the mathematician's testimony).

Matt Kuenning said...

Hi Carrie,

You raise some good questions. Let me try again.

We're looking for a non-question-begging reason for the common view that sense experience can only improve our epistemic position regarding what's so, not what's possibly or necessarily so. My proposal is that, for every non-modal p, the relevant modalities are part of the conditions for sense experience to be evidence for p. Therefore the experiential evidence for p doesn't transmit to the modality if we infer the modality from p.

First you ask why we would have to infer the modal from p instead of just taking the experience as direct evidence for the modal truth. I'd say that if there's no causal connection between the modal fact and my sense organs, I need an inference to make sensory evidence into evidence for a modal. Plus, I can't think of a plausible example of direct sense evidence for a modality. You give the example of hearing testimony from a mathematician as a case where the evidence for p and possibly p arrive at the same time. I'm not sure how to explain why testimony is evidence, but I'm pretty sure the explanation is not that testimony simply reduces to sense evidence so I'd conclude that's not an example of direct sense evidence for a modal.

You also ask why we should think it's necessary that we can't have sense evidence for p unless we're already entitled to accept possibly p. I think our relation to the relevant modals isn't something we have to know or justifiably believe. Clearly you're right that I can have sense evidence for p without ever having thought that possibly p. So the proposal is going to have to be that I have to be entitled to possibly p, or permitted to accept p, or be allowed to take possibly p for granted, to get sense evidence for p. As you pointed out, the problem then becomes: what's the (non-question-begging) motivation for that thought? One motivation I mentioned is that it could explain the intuition that sense experience can't be evidence for or against mathematical truths. But that doesn't seem to get us very far, even if those intuitions are widely shared, because there could be other explanations and nothing so far indicates warrant transmission failure is the best explanation.

I think if my proposal can be motivated, it will have to be by considering what would have to be the case if one weren't entitled to possibly p, or permitted to take possibly p for granted. Remember, p is the sort of proposition that sense experience can directly warrant. If possibly p is true, then my sense is that I'd be entitled to possibly p unless I've done something or been put in a situation such that if I thought about possibly p, it would seem false to me. For example, possibly there are black swans, so I'm entitled to assume it's possible, even if I've never thought about it, unless something has messed up my entitlement. What could do that? I think it would have to be that I'm in the grip of some bizarre theory that being white is essential to being a swan, or I've become convinced that only minds exist and the external world is an illusion. If I thought something like that, then would seeing a black swan count as evidence for me that there's a black swan? My intuition is no.

I'm not sure if this is any better than my first attempt, but I'd be interested to hear your thoughts.

Carrie Jenkins said...

Hi Matt,

"I need an inference to make sensory evidence into evidence for a modal."

Even if that's right, and even assuming that the inference has to go via some modal proposition, presumably the inference could goes via some *other* modal than the one that we're inferring to. So isn't there room for a form of non-direct empirical grounding for (at least some) modal propositions which is not subject to transmission of warrant worries?

Also, I wonder what sort of notion of entitlement you're using, according to which you are entitled to something provided it's true and there's nothing messing up the entitlement? This isn't how Wright thinks entitlement works; but perhaps you had something else in mind?

If all that's required for entitlement is truth and nothing getting in the way, though, then I'm not sure we can motivate a transmission of warrant worry even if it should turn out that entitlement in *this* sense to 'possibly p' is required before empirical grouding for 'possibly p'.

Basically, it starts to look like entitlement is not really an epistemic notion. But you'd need it to be epistemic in order to argue that there's something wrong with using empirical evidence to ground 'possibly p' because in order to appreciate that it is evidence we require prior 'entitlement' to 'possibly p'. If all that means is that it's required in advance that 'possibly p' be true and there be nothing getting in the way, that doesn't appear to interfere with our use of the evidence to ground 'possibly p'.

Daniel Elstein said...

Carrie, going back to your question about why 'Water is necessarily H2O' isn't a good explanation of 'All water is H2O'. First let's distinguish: we aren't trying to explain why all the watery stuff is H20 - the explanation of that is presumably either that no other compound is watery, or that other watery compounds are extremely unlikely to arise. 'Water is necessarily H20' means (roughly) 'The watery stuff of the same natural kind as the watery stuff which our actual linguistic community uses as an exemplar of "water" is necessarily H20', and 'All water is H20' means 'All the watery stuff of the same natural kind as the watery stuff which our actual linguistic community uses as an exemplar of "water" is H20'. Now the former is entailed by the latter, because of the rigidity imparted by the term 'actual'. We know that water is necessarily H20 because we know that all water is H20, we know the semantics of 'all water is H20', and we know some basic modal logic. So what's wrong with the explanatory claim you suggest is that the direction of explanation goes the other way. Now I suppose that there might be some people who don't know *that* all water is H20, because they don't know that 'water' is a rigid designator. But then what they lack is semantic knowledge, not modal knowledge, and without that semantic knowledge they just won't believe that water is necessarily H20. Someone who asks *why* all water is H20 presumably really wants to know why all the watery stuff is H20, but that isn't explained by water necessarily being H20.

I'm afraid that all of this turns on how we cash out what Putnam says about the meaning of 'water', and I suspect that what I say isn't orthodox. But basically I'm puzzled as to what kind of person could find 'water is necessarily H20' a good explanation of all water being H20. Either they know that 'water' is a rigid designator and what that entails, or they don't. Either way, it doesn't look like a good explanation.

Brad Majors said...

Hi Carrie,

Thanks for your response. You write: 'But I'm worried that what comes after the 'because' here looks to me a bit too much like a restatement of the thing it's supposed to explain. Unless I'm already convinced that experience cannot deliver modal knowledge, why should I accept that "experience can tell us only how things are in this world"?'

I think the answer to this turns on the connection between experiential knowledge and causation. Causal relations -- or at least *most* causal relations -- are plausibly taken to be metaphysically contingent. And I think that perceptual knowledge (e.g.) typically requires, as part of its epistemic warrant, that these relations obtain. If this is correct then it would seem to follow that there can be no perceptually grounded knowledge of metaphysical necessities.

[There is a reasonably clear discussion of closely related points in Peacocke's _Realm of Reason_, pp.242-3.]

Carrie Jenkins said...

Daniel,

"We know that water is necessarily H20 because we know that all water is H20, we know the semantics of 'all water is H20', and we know some basic modal logic. So what's wrong with the explanatory claim you suggest is that the direction of explanation goes the other way."

This sounds like you think the *direction of epistemic access* goes the other way (from the non-modal claim to the modal one). But surely this is normal in cases of inference to the best explanation - the direction of explanation goes one way (from the rain to the wet streets), the direction of epistemic access goes the other way (from the wet streets to the rain).


Brad,

"I think that perceptual knowledge (e.g.) typically requires, as part of its epistemic warrant, that [causal] relations obtain."

I'm sure that something like this assumption is what underwrites most people's intuition that experience can't give us epistemic access to modal truths. But when it's brought out, I tend not to feel compelled to agree with it! The motivating question seems to be: how could experience give you knowledge that p unless p were appropriately causally related to the things you experience? And I want to answer: as long as p is the best explanation of what you experience (and assuming that some explanations are non-causal), inference to the best explanation could enable you to have experientially grounded knowledge of p even when p's not causally connected to the things you experience.

Matt Kuenning said...

Hi Carrie,

If I'm counting right, so far we've got two proposed examples of modal knowledge built on sense experience: 1. We know empirically that all water is h2o, and we infer abductively that its necessarily so. 2. We know empirically there is a black swan and infer that black swans are possible. I'm still convinced these aren't genuine examples because the modal conclusion underwrites (in some yet-to-be-adequately-explained way) the ability of the sense experience to be evidence in the first place. But let me switch sides and offer a more complicated example that strikes me as possibly genuine.

Suppose we believe that a moving body necessarily loses momentum - that it's somehow essential to motion that it costs energy. Galileo starts explaining motion based on the proposition that in at least one possible world - the actual one - a moving body doesn't lose momentum unless something interferes. We get empirical evidence that Galileo is right. Isn't that also empirical evidence for us that it's not necessarily so that motion costs engergy?

If the answer is yes, maybe this shows you were right earlier when you said that even if some modal truths have to be part of our entitlement for sense experience to be evidence,it doesn't follow that the particular modal truth that we want to base on experience had to be part of that set.

Daniel Elstein said...

Carrie, I take your point: direction of epistemic access is often (usually?) opposite to direction of explanation. However, the case we're talking about is one where the alleged explanandum entails (given linguistic conventions) the alleged explanans. The typical cases where the directions of epistemic access and explanation are opposite are where the inference from the explanandum to the explanans is defeasible. I know of lots of cases of explanation where the explanans entails the explanandum (take the hypothetico-deductive model), but none where the explanandum entails the explanans (but perhaps there's a neat counterexample that I'm missing). For instance, the area of the rug being 20 sq ft is explained by the rug being 4' by 5', and here the explanans entails the explanandum but not vice-versa. I erred if I suggested that what defeated your diagnosis of explanation was that we know that water is necessarily H2O via knowing that all water is H2O; what I should have emphasised is that that route to knowledge of water necessarily being H2O is deductive.

Carrie Jenkins said...

Matt,

Can you say a bit more about how your new case differs from the black swan example? There we had empirical evidence for 'there is a black swan' and inferred to 'possibly there is a black swan'. Here we have (something close to) the following: empirical evidence for 'motion doesn't cost energy' and an inference to 'it's not necessary that motion costs energy'. But if the swan case is an inference from p to possibly p, and the new case is an inference from q to not-necessarily-not-q, then assuming that 'possibly' and 'necessarily' are interdefinable in the usual way, surely the two cases should be equally plausible?


Daniel,

The existence of a deductive route of the kind you describe from 'all water is H20' to 'necessarily all water is H2O' isn't something I would query. But why does the existence of this deductive route make it impossible for there to be an abductive route *as well*? Why not let a thousand modes of epistemic access bloom? Well, two anyway! :)

Matt Kuenning said...

Hi Carrie,

As you point out, the Galileo case and the black swan case are the same at the most general level of description. Here are two differences. First, in the former case there is a genuine dispute about the modal conclusion before Galileo gets proved right. Second, the most general description of the black swan case is pretty much all there is to it. I get empirical evidence supporting p and then straight off infer possibly p. Galileo's case is more complicated. I get empirical evidence for some observable facts x, y and z, which support a scientific theory that includes p, and from p we infer the modal (and there is a point in doing so because of the dispute).

If you've got the intuition, as I do, that the black swan case is a spurious example knowing a modal truth based on sense experience, but also think that the Galileo case may well be a genuine example, then there is an interesting question explaining how the added features of the Galileo case make a difference. Do you think the added features of the Galileo case couldn't make a difference as to whether the reasoning gives us epistemic access to the modal? In other words, do you think that even if one did have diverging intuitions on the two case, that fact couldnt' tell us anything?

My family and I are leaving on vacation later today, and my computer access will be unpredictable. In case this thread dies out and you move on to something else, I want to say thanks for the discussion and for the blog in general.

Daniel Elstein said...

Carrie, we agree then that 'All water is H2O' and 'Water is necessarily H2O' are deductively equivalent, i.e. they are deductive consequences of each other (I'm inclined to say that they're logically equivalent, but I wouldn't want to fight over it). They certainly have the same truth conditions. If I know that P and Q have the same truth conditions, I'm never going to go from one to the other by abduction. So in order for it to make sense for me to use abduction to get from one to the other, I have to be ignorant of at least one of their truth-conditions (or perhaps I somehow know both, but fail to see that they're the same). But if you have to be ignorant of the truth conditions of P and Q in order to use abduction to get from one to the other, how is that abduction supposed to work? How would I know that Q explains P if I don't know what it takes for one of them to be true? That suggests to me that there's something you need to say to make that kind of abduction plausible.

Here's another reason why these won't be good candidates for abduction. I'll concede arguendo that there can be explanations between equivalent sentences - maybe the lines in a proof explain the conclusion. But I don't see how such explanations can be the 'best' explanations of their explananda - the criteria for the goodness of such explanations are relative to the intellectual capabilities of the audiences. Suppose you say that water necessarily being H2O is the best explanation of all water being H2O, and I say that 'water' referring rigidly to H2O is the best explanation; what decides which of us is right?

More controversially, I'd say that genuine explanations are between states of affairs or facts (or maybe propositions), not sentences. The kind of explanations that can occur between different sentences pointing to the same fact are not candidates for abduction, which is a way of finding out facts which we didn't already know. 'Water is necessarily H2O' and 'All water is H2O' point to the very same fact, so I think that they can only seem to explain each other. In the cases where we call proofs explanations, we should really talk about whether they are good or bad *demonstrations*. But maybe you have rather different intuitions.

Carrie Jenkins said...

Matt,

I don't have a different intuitive response to the two cases, but I think that if I did I'd be hard pressed to say what was *relevantly* different about them and how it made the difference.


Daniel,

I don't think I'd be happy to go as far as saying that 'All water is H2O' and 'Necessarily all water is H2O' have the same truth conditions or point to the same fact. The truth condition (relevant fact) for the second thing involves all possible worlds, but the truth condition (relevant fact) for the first thing doesn't.

What I would agree is that the first thing is true in exactly the same circumstances as the second thing (i.e. they're both true at all possible worlds). Is your underyling intuition that *all* necessary truths have the same truth condition and point to the same fact, and that therefore there can't be relations of explanatory dependence between them? If so, I think that may be the point at which our intuitions differ.

Daniel Elstein said...

Carrie, as I said I'm happy to admit that there is some kind of explanatory dependence between necessary truths. The fact that 14 has a prime divisor (which isn't itself) explains why 14 isn't prime. And I guess that even pretty fine-grained theories of facts are going to say that '14 has a prime divisor (which isn't itself)' and '14 isn't prime' point to the same fact. Nevertheless, I'd find it very odd if someone told me that they'd discovered *by abduction* from 14 not being prime that 14 had a prime divisor other than itself. (I wonder if you have the same intuition about that case?) My conjecture above was that the reason why it's possible to admit there's an explanation in such a case without there being any possibility of abduction is that we can't give any sense to the idea that one necessary truth is the *best* explanation of another, and I went on to say that it seemed to follow that we should say that such explanations aren't genuine (but that last point isn't really crucial). Here's a challenge. Suppose you share my intuitions about the prime case. How would you explain why there can be explanations without any hope of abduction arising from them?

You might want to argue that there can't be abduction when the truths do point to the same fact, but that there can be abduction between necessary truths that don't point to the same fact. Now I'd want to concede something here, which is that there might be abduction between some necessary truths, e.g. 'Water is H2O' and 'Dry ice is CO2'; what I think marks out such cases is that there is no deductive (a priori?) route between the two truths. I gave a justification for this: no deductive consequence of P can be the best explanation of P. In virtue of what could one deductive consequence of P be a better explanation of it than another?

I think that the point I'm making is independent of one's theory of facts, and is also compatible with saying that there can be explanatory relationships between necessary truths.

Carrie Jenkins said...

Daniel,

'what I think marks out such cases is that there is no deductive (a priori?) route between the two truths.'

What would you say about cases where there is a deductive route but it is not known? Suppose we know all water is H2O but we haven't done enough phil of language to know about rigid designation and the like. So we don't have access to the deductive route from the universal claim to the modal claim. Still, mightn't we posit the modal claim as the best explanation of the universal? (A claim about the rigidity of 'water' wouldn't be such a good explanation at this stage because we don't know what rigidity is.)

(Something I should think about, even if that's plausible, is what makes the difference between cases where we have a universal claim which clearly isn't best explained by the corresponding modal claim and the (putative) cases where this sort of explanation could look promising.)

Carrie Jenkins said...

More on that last point (I'm now responding to my own comments! That didn't take long!):

It's not too controversial to think that claims of *nomological* necessity can be known abductively, and essentially the same issue (the question of when to treat a regularity as law-like) comes up there too.

Another question: if we are happy with abduction to claims of nomological necessity but not to claims of metaphysical necessity, what's the relevant difference?

Daniel Elstein said...

Carrie, I think these are the right cases to think about. On the one where we don't know about rigidity, there seem to be two problems with the proposed abduction. Firstly, I think that 'best explanation' isn't relativized to what is known (though perhaps 'explanation' is). So if the existence of the deductive route prevents the explanation being best, ignorance of the deductive route doesn't change that. Putting it a little differently, if the rigidity of 'water' is the actual best explanation of all water being H2O, my ignorance is no defence. As I understand it, abduction will only create knowledge if the explanation abduced is in fact the best explanation; thinking that it's the best explanation may make it a reasonable piece of abudction, but it won't create knowledge (maybe it's a Gettier case). Secondly, if you don't understand that 'water' is rigid, you won't know what it is to say 'all water is H2O' - you're likely to think that it means that all watery stuff is H2O. But that isn't explained (even if it's true) by the fact that water is necessarily H2O. So I imagine you want the case where we know *implicitly* that 'water' is rigid, though not explicitly. But I don't see how we can know it implicitly without the deductive route being available. Of course, we could fail to take the deductive route for other reasons of incompetence, but you're presumably not arguing that the abduction is ok in that kind of case.

As for nomological necessities, that's a very involved issue because it probably depends on our views of laws of nature. But here's my first shot at what to say (maybe I've just missed the problem). If you think that an event is explained by its nomological necessity, that can be cashed out in two ways. Either it's simply because all events are determined by laws of nature (and initial conditions) in which case the inference is a priori, or you think that there a particular law of nature which explains the event, in which case the abductive inference is to the law itself, not to the fact of nomological necessity. Of course I admit that nomological necessities are known a posteriori, just not abductively.

Carrie Jenkins said...

Daniel,

Didn't your earlier argument depend on our accepting that "the criteria for the goodness of such explanations are relative to the intellectual capabilities of the audiences"?

Also, I thought that your reason for saying that the deductive route interferes with the abduction
is that the former always renders the latter obsolete ("If I know that P and Q have the same truth conditions, I'm never going to go from one to the other by abduction"). But that's not the same as showing it's *not the best explanation*, as you now seem to be assuming. Our grasp of the deductive route only means that we *don't need to use* the abductive step, not that the abductive step would be a bad one.

Daniel Elstein said...

Carrie, I realise I'm being slightly confusing on those points, sorry! I think there is a dimension of goodness which is relative to intellectual capabilities, which we can think of as how enlightening the explanation is. I accept that there can be enlightening explanations between truths that entail each other. But my intuition is that whatever goes into making an explanation 'best' from the point of view of abduction can't just be how enlightening the explanation is to an audience, otherwise the knowledge that can gained through abduction will be relative in (what seems to me) a bad way. So I have to think that there are other ways an explanation can be good that aren't themselves relative to intellectual abilities.

On the other point, I don't think I said that it was the redundancy of abduction that made it bad; I just can't see how a deductive consequence of P can be the best explanation of P. That's my intuition; I thought I had some arguments for it too. One was rhetorical: "In virtue of what could one deductive consequence of P be a better explanation of it than another?" (I should add - suppose we're talking about the deductive consequences of P that also entail P). If it's just how enlightening the explanation is, then we'll get the bad relativity: I can know Q by abduction from P only if I happen to find Q an enlightening explanation of P.

Perhaps it's getting to the stage where I'm just reeling out a bunch of intuitions bearing no relation to yours; in which case, sorry.

Carrie Jenkins said...

Daniel,

No apology necessary! It's been interesting to work through. I don't have the intuition that no deductive consequence of p can be the best explanation of p. I guess I think even logical truths might explain each other, although they're obviously logically equivalent. E.g. perhaps the best explanation of why:
(1)the cat is on the mat --> (the mouse is in the house --> the cat is on the mat)
is that:
(2) if the consequent of a material conditional is true the whole conditional is true.

On the other hand, neither do I have a well-worked-out account of explanation of this kind to offer, which would answer your rhetorical question. All I can do is point to instances: the explanation described above is better than an attempt to explain (1) by stating some irrelevant logical truth, e.g.:
(3)p&q --> p.

Another intuition I have that you don't is that explanatory 'bestness' is relative to audience. I can't get a grip on any idea of explanatory goodness that isn't so relativised. (Besides, surely there are lots of other ways to explain what makes the difference between merely reasonable abduction and knowledge-generating abduction, depending on your theory of what knowledge is in general. So it's not obvious that we need 'absolute bestness' to fulfil that function. E.g. you might say that abduction is knowledge-generating in those kinds of cases where it is reliable, and not otherwise.)

So I guess those are the two points of disagreement!

Daniel Elstein said...

Carrie, I agree that there can be a best explanation of a logical truth for a particular person. If that's all the material conditional example is meant to show, then fine. But I imagine that for some people a better explanation will be:
(3) a material conditional is only false when its antecedent is true.
And for some people the conjunction of (2) and (3) will be best, etc.

As I understand it, your account of the relationship of abduction to knowledge goes like this:
Abduction is inferring the (subjectively) best explanation. Abduction generates knowledge iff (roughly) the subject is set up so that inferring what is for her the best explanation is a reliable method.

And my account goes more like this:
Abduction is inferring the objectively best explanation (so if the explanation you infer is subjectively but not objectively best, you have abduced unsuccessfully, though perhaps reasonably). (Successful) Abduction generates knowledge iff what is abduced is true.

One reason for favouring an objective conception of best explanation is that people argue about what the best explanations of various phenomena are. They wouldn't if they were just talking about which explanation was subjectively best. Now I see that it would be silly to argue about the best explanation of logical truths, but that is evidence (for me) that abduction isn't applicable to logical truths.

Carrie Jenkins said...

"One reason for favouring an objective conception of best explanation is that people argue about what the best explanations of various phenomena are. They wouldn't if they were just talking about which explanation was subjectively best."

They might, if they thought their subjective positions were identical in all relevant respects and each thought the other was wrong about which was the best explanation for someone in that position.

Also (just for the record!) I didn't (and don't) endorse the reliabilist proposal - that was just illustrative. My own preferred account of what makes for knowledge is explanationist rather than reliabilist.