Wolfgang Schwarz

As I read "Elusive Knowledge," perceptual experience and memory are individuated by their content. That is, worlds in which it does not perceptually appear to me that there is a table in front of me are eliminated, and worlds in which I do not seem to remember that I had cereal for breakfast are eliminated. (Of course, worlds in which it false that there is a table in front of me, or that I had cereal for breakfast, are not thereby eliminated.)

The content can be modelled as a class of centered possible worlds. There need not be any sentence in one's language that expresses that self-locating proposition.

It is then a consequence of Lewis's account of knowledge that we always know what the content of our perception and memory is. Many externalists would of course find that objectionable, but Lewis had independent reasons to to posit an internalist, narrow kind of content.

If some sort of functionalism is assumed to be necessarily true, then there is not much to choose between that reading and yours. A world in which it perceptually appears that p will be a world where I am in such-and-such causal/functional state. Assuming the necessity of functionalist analyses (I don't know whether Lewis was committed to that), I think your criticism stands: Lewis's account falsely implies that we have knowledge about our causal organization from the armchair (in fact, even those of us who say that they do not believe in functionalism!). An account that entails modal omniscience seems especially problematic if it also entails omniscience about one's intentional content.

Lewis says that everyone always knows the necessary proposition, although of course we will often fail to recognize it in various disguises. I am not sure whether this is what one wishes to say about the necessary proposition, but if you are willing to grant Lewis this, then I do not see that the examples discussed here present any additional problem; presumably the apparent failures to know similarly involve failing to recognize knowledge in various disguises. On the other hand, if you're not willing to grant Lewis that, what need for the additional examples?

I once had a similar idea. Doesn't Lewis' definition of knowledge let us know our qualia (whether physical or non-physical)?
I think I understand Wo's point (or do I read my own idea into it?), but I don't understand Stephans reply. I recall that Lewis is rather explicit in EK that experiences be individuated by intrinsic nature rather than by intentional content.
My own idea was that the rule of belief will bring relief. Normally you have to belief what you know. And you have no beliefs that determine, by their primary intensions, the intrinsic nature of your states of experience. By their intrinsic nature your experiences have the discriminative power to distinguish between words where you have experiences of THIS very kind from worlds where you have experiences of different kind (translate this properly into centred worlds talk). But this discriminating power does not endow you with additional knowledge, because you cannot believe what you would know if you were able to believe it.
But the rule of belief is not strict, and I'm not sure whether Lewis' characterisation excludes that there are exceptions that would yield knowledge of qualia. But the exceptions he considers are of a very different kind.
Wo: hope you are fine.

Hi all, and thanks!

Re Stephan's proposal: Lewis is explicit that the possibilities eliminated by an experience E are not the possibilities eliminated by E's content, not even by a very narrow content of the kind: it seems to me that there is a table in front of me. But maybe Stephan's suggestion is that E is nevertheless individuated by its content. The view would be, as Lewis puts it on top of p.425, that E is "in some sense, fully characterized by [its content] P". That would amount to some kind of pure representationalism, which seems weird to me. Though Lewis says he takes it to be "an open possibility". I have to think about what this would mean for various other things Lewis holds.

I agree with Stephan that this position in the end comes down to something very similar to what I proposed in the posting, giving us knowledge of our causal structure from the armchair.

Ralf: On the face of it, it seems that the rule of belief couldn't help: All the rules in EK tell us how the quantifier in "S knows that P iff P holds in every possibility left uneliminated by S's evidence" is restricted. But if P is the proposition that I'm in experiential state E, then P holds in absolutely every uneliminated possibility; therefore it also holds in all of its subsets, no matter by what rules they are restricted, right? On the other hand, you're right that Lewis's reasons for allowing knowledge without belief are of a very different kind, and perhaps this is another reason to turn from a purely evidential theory of knowledge to one that has a tighter connection to belief. A rule for proper ignoring doesn't seem to do that though.

And yep, fine. Hope you are too!

Aaron: You're right that one might say we really do know that we are in brain state X, but just fail to recognize that proposition, like we often fail to recognize the necessary proposition. But I think there are important differences between the cases. For instance, it is clear that we know the necessary proposition under *some* of its disguises, but it is far from clear that I know the proposition that I'm in brain state X under any disguise (going by primary intension). Moreover, it seems more absurd to me that we can knowledgably rule out scenarios in which we have no brains than that we can knowledgably rule out scenarios in which some necessary truth is false (since there are no such scenarios). Finally, when Lewis rejects the 'identification thesis', he appears to explicitly reject the view that we know in what experiential states we are. So I don't think this answer is applicable here.

I'm afraid Dr. Wo is right about the rule of belief, so I'm puzzled again. I've always taken it for granted that by "experience" Lewis means the intrinsic nature of our states of experience, not their playing a certain topic-neutral folk-psychological role. Remember that specific input-output connections between my inner states and my environment belong to those roles. And remember that the roles are defined as the typical roles the states play in my species. So I can absolutely exclude worlds where (or alternatives of mine with respect to which) those input-output connections do not obtain, e.g. where I'm a brain in a vat? I can absolutely exclude worlds where I'm not a member of a natural species? I can absolutely exclude worlds where there is no causation, where no laws of nature obtain? This doesn't sound like Elusive Knowledge at all.
Desperate question: how obvious is it that by the propositions P we are supposed to know Lewis means primary intensions? I may well know the proposition that I'm in such-and-such a brain state, which is my experience, in the sense of a secondary intension.

Hmm! You're right, it would be bad if the functional role account meant we can exclude brain in a vat possibilities. Might it help that the roles are narrow in the sense that our twins on twin earth and us have states that play exactly the same role? Recall Lewis's discussion in "Reduction of Mind": a ball-kicking desire doesn't normally cause ball-kickings; rather, it causes behaviour such that if there was a ball in front of you and things were otherwise as we think they normally are, then the impact of your foot would dislocate the ball. Likewise, an elephant-experience arguably need not be caused by elephants; rather, its cause should be such that on the assumption that things are as we think they normally are, it implies the presence of elephants. Or so. Admittedly, it's not clear that this could apply not only to our normal twins, but also to our envatted twins who for instance don't behave much at all.

An alternative would be to treat the brains in a vat as madmen, in the sense of "mad pain": their states are what they are due to their role in normal members of their kind. On the role account, what our experiences tell us is that we are in a state that typically plays such-and-such a role in normal members of our kind. This doesn't rule out that we are non-normal envatted members.

It does however rule out that all members of our kind are envatted, and it still seems strange that we should be able to rule that out for sure. Though, like with your species worry, one should remember that the relevant "kind" need not be a natural species, and might only consist of us and some of our counterparts. Not that this helps very much: how would normal people end up being counterparts of a lonely brain in a vat? Appeal to sameness of mental content would be circular.

Still, there's hope that we can avoid the bizarre consequence that every experience establishes as certain that we are not brains in a vat, and there are other people in the world.

Re causation and laws, maybe that's not so bad: could there really be a world without laws or without causation? On Lewis's analyses, that's not clear at all. If there are no such worlds, it wouldn't be a problem that we can immediately rule them out.

There remains the problem that every experience would tell us for certain that folk psychology is largely correct. That seems crazy given that folk psychology is an empirical theory (and a strong theory, as Lewis insists).

But perhaps one can break up the bundle: If it turns out that some fragment of folk psychology -- say, about dreams -- is completely wrong, that wouldn't imply that all folk psychological terms suddenly become empty, that there are no such things as belief and pain. The effect would be limited to those parts of folk psychology that are close to the relevant fragment. Other parts would survive by satisficing. Similarly, if some kind of evidence strongly supports only a small fragment of folk psychology, it needn't thereby put the whole theory beyond doubt. Having an experience might then only guarantee that a small fragment of folk psychology is true. And perhaps that's not too absurd.

I think Armstrong argues in "A materialist theory of the mind" that when we recognize one of our states as pain, we thereby recognize it as something that should be avoided. The evidence that we're in pain would therefore tell us that we're in a state that people like us normally tend to avoid.

On you desperate question: if I understand you correctly, that was basically Aaron's proposal. I don't think it goes well with what Lewis says. But maybe I don't understand you correctly? Let P be the set of worlds where I'm in brain state E. Let P+ be a superset of P that is the 1- and 2-intension of some semantically stable sentence S, such as perhaps the the set of worlds where I have a brain (with S = "I have a brain"). It seems that "I know for sure that S" is still false. And this time, it doesn't matter if we go by 1- or 2-intension.

Wo: "a ball-kicking desire doesn't normally cause ball-kickings; rather, it causes behaviour such that if there was a ball in front of you and things were otherwise as we think they normally are, then the impact of your foot would dislocate the ball."
How is this supposed to help? A internal state/output connection that captures the output as some behaviour that has such-and-such a counterfactual profile seems bad enough. The profile is a very worldly one: one that concerns the movement of balls.
I would still feel very uneasy about the assumption of an absolute ability to exclude worlds where my states don't play the folk-psychological role they actually play.

laws and causation: This is interesting, I would have thought that Lewis clearly assumes lawless worlds. Doesn't recombination give us worlds where our fundamental properties - the masses, charges etc. - are distributed in a completely chaotic way? How could there be any guarantee that for any possible recombination there is an axiomatic system that is distinguished as providing the best balance between strength and simplicity?

One important point might be that Lewis explicitly only analyses "s knows proposition P", not "s knows that S". So his analysans is rather artificial from the very beginning. It is not quite our ordinary concept of knowing-that. Perhaps he just admits that according to his analysis it turns out that we know a proposition P that is neither the primary nor the secondary intensions of any of our sentences, so that this case of knowledge cannot be attributed in the form "s knows that S".

He might then further say that "s knows that S" is ambiguous between attributing knowledge of the 1-intension of "S" or the 2-intension of "S". Depending on how a particular attribution is meant, it turns out true or false.

If there happens to be a sentence "S" that has the proposition P = *I am in E* as its 2-intension, then "I know that S" turns out true on the second reading, where the 2-intension of "S" is the relevant one. But there is no sentence "S" that expresses P, i.e. has P as its 1-intension.

Hoi! The counterfactual proposal was supposed to provide a way of specifying the input-output connections in such a way that even a brain in a vat could have them -- so that we needn't say that our evidence excludes brain in a vat situations.

Not sure about the laws. Does "best theory" entail "good theory"? Even in a totally chaotic world, some theories will be better than others, for instance, true theories will be better than false ones, and theories that assign high probability to things that actually happen will be better than theories that always assign low probability to actual events. There certainly won't be a unique best theory. But in this case, I thought the laws would be indeterminate. Moreover, even if there are worlds without laws, it is a further question whether there are worlds without causation. On Lewis's counterfactual analysis, I don't see how this is possible. The miracle clauses for similarity of course become empty, but the two other clauses remain.

I agree about Lewis analyzing a somewhat technical notion of propositional knowing. But surely this notion should bear some relation to our ordinary notion. And saying that we always know for certain in what brain states we are seems inacceptable, especially if it can't be explained away, like our apparent ignorance of mathematical truths. Re your last proposal, wouldn't that still mean that "I know that S" is determinately true for the strongest semantically neutral (1-intension = 2-intension) sentence that expresses a superset of the proposition that I'm in brain state E -- perhaps "I know that I have a brain"?

I remain seriously puzzled by all this, because the question seems so obvious that either Lewis would have noticed it himself and given an answer, or the answer is so trivial that Lewis never even bothered about the question. And I still tend to think that the consequences of the view that an experience is individuated by its intrinsic nature are less repugnant than the consequencesn of the view that it is individuated by folk-psychological role.

How exactly is the brain in the vat supposed to exhibit an outer behaviour that has the ball-kicking counterfactual profile?

Suppose my momentary actual experiential state E involves exactly 6-10^23 ontically atomic particles over which 11 monadic perfectly natural properties are distributed. Sentence S = "I am in a state that involves exactly 6-10^23 ontically atomic particles over which 11 monadic perfectly natural properties are distributed" seems to be a perfectly accessible sentence. By being in E I epistemically exclude all altermnatives where I'm not in E. Hence I exclude all alternatives where I'm not in a state which etc. Hence I absolutely know the egocentric proposition that I am in a state etc. And this is a proposition I can express by sentence S. Moreover, it can be attributed to me by embedding S or a more or less trivial variant of S in a that-clause: "Ralf knows that he is in a state etc." Such attributions would be true, whether or not I have any justification for believing that I am in a state etc., even whether or not I believe or tend to believe this. One would even say that I should rather not believe this.
Is this the point? It would indeed be quite embarrassing if nothing more could be said about this.

Concerning laws I would say that even when we admit worlds where there is no system of or including regularities that turns out *robustly* best, it is an open question whether *any* axiomatic system is distinguished as one best balance between simpicity and strength.
Concerning causation: I do not see that the rest of the counterfactual analysis of causation provides anything of interest once the reference to laws drops out. What connects the closest-possible absences of c with absences of e if not laws of nature?